Update generated Python Op docs.

Change: 138805538

12 files changed, 1830 insertions, 2 deletions

diff --git a/tensorflow/g3doc/api_docs/python/contrib.distributions.bijector.md b/tensorflow/g3doc/api_docs/python/contrib.distributions.bijector.md
index d2d02ec326..c7d4d68f3e 100644
--- a/tensorflow/g3doc/api_docs/python/contrib.distributions.bijector.md
+++ b/tensorflow/g3doc/api_docs/python/contrib.distributions.bijector.md
@@ -163,6 +163,10 @@ Subclass Requirements:
   functions to avoid computing the `inverse_log_det_jacobian` or the
   `inverse`, respectively.
 
+- Subclasses should implement `_get_forward_event_shape`,
+  `_forward_event_shape` (and `inverse` counterparts) if the transformation is
+  shape-changing.  By default the event-shape is assumed unchanged from input.
+
 Tips for implementing `_inverse` and `_inverse_log_det_jacobian`:
 
 - As case 3 [above] indicates, under some circumstances the inverse function
@@ -280,6 +284,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.forward_event_shape(input_shape, name='forward_event_shape')` {#Bijector.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Bijector.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -306,6 +330,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.get_forward_event_shape(input_shape)` {#Bijector.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Bijector.get_inverse_event_shape(output_shape)` {#Bijector.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.inverse(y, name='inverse', **condition_kwargs)` {#Bijector.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -363,6 +429,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Bijector.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Bijector.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -531,6 +617,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.forward_event_shape(input_shape, name='forward_event_shape')` {#Chain.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Chain.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -557,6 +663,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.get_forward_event_shape(input_shape)` {#Chain.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Chain.get_inverse_event_shape(output_shape)` {#Chain.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.inverse(y, name='inverse', **condition_kwargs)` {#Chain.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -614,6 +762,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Chain.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Chain.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -759,6 +927,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.forward_event_shape(input_shape, name='forward_event_shape')` {#CholeskyOuterProduct.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#CholeskyOuterProduct.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -785,6 +973,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.get_forward_event_shape(input_shape)` {#CholeskyOuterProduct.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.get_inverse_event_shape(output_shape)` {#CholeskyOuterProduct.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse(y, name='inverse', **condition_kwargs)` {#CholeskyOuterProduct.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -842,6 +1072,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse_event_shape(output_shape, name='inverse_event_shape')` {#CholeskyOuterProduct.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#CholeskyOuterProduct.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -984,6 +1234,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.forward_event_shape(input_shape, name='forward_event_shape')` {#Exp.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Exp.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -1010,6 +1280,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.get_forward_event_shape(input_shape)` {#Exp.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Exp.get_inverse_event_shape(output_shape)` {#Exp.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.inverse(y, name='inverse', **condition_kwargs)` {#Exp.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -1067,6 +1379,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Exp.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Exp.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -1194,6 +1526,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.forward_event_shape(input_shape, name='forward_event_shape')` {#Identity.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Identity.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -1220,6 +1572,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.get_forward_event_shape(input_shape)` {#Identity.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Identity.get_inverse_event_shape(output_shape)` {#Identity.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.inverse(y, name='inverse', **condition_kwargs)` {#Identity.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -1277,6 +1671,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Identity.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Identity.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -1367,7 +1781,7 @@ exp = Inline(
 The above example is equivalent to the `Bijector` `Exp(event_ndims=1)`.
 - - -
 
-#### `tf.contrib.distributions.bijector.Inline.__init__(forward_fn=None, inverse_fn=None, inverse_log_det_jacobian_fn=None, forward_log_det_jacobian_fn=None, is_constant_jacobian=False, validate_args=False, name='inline')` {#Inline.__init__}
+#### `tf.contrib.distributions.bijector.Inline.__init__(forward_fn=None, inverse_fn=None, inverse_log_det_jacobian_fn=None, forward_log_det_jacobian_fn=None, get_forward_event_shape_fn=None, forward_event_shape_fn=None, get_inverse_event_shape_fn=None, inverse_event_shape_fn=None, is_constant_jacobian=False, validate_args=False, name='inline')` {#Inline.__init__}
 
 Creates a `Bijector` from callables.
 
@@ -1380,6 +1794,14 @@ Creates a `Bijector` from callables.
     log o det o jacobian of the inverse transformation.
 *  <b>`forward_log_det_jacobian_fn`</b>: Python callable implementing the
     log o det o jacobian of the forward transformation.
+*  <b>`get_forward_event_shape_fn`</b>: Python callable implementing non-identical
+    static event shape changes. Default: shape is assumed unchanged.
+*  <b>`forward_event_shape_fn`</b>: Python callable implementing non-identical event
+    shape changes. Default: shape is assumed unchanged.
+*  <b>`get_inverse_event_shape_fn`</b>: Python callable implementing non-identical
+    static event shape changes. Default: shape is assumed unchanged.
+*  <b>`inverse_event_shape_fn`</b>: Python callable implementing non-identical event
+    shape changes. Default: shape is assumed unchanged.
 *  <b>`is_constant_jacobian`</b>: `Boolean` indicating that the Jacobian is constant
     for all input arguments.
 *  <b>`validate_args`</b>: `Boolean` indicating whether arguments should be checked
@@ -1421,6 +1843,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.forward_event_shape(input_shape, name='forward_event_shape')` {#Inline.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Inline.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -1447,6 +1889,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.get_forward_event_shape(input_shape)` {#Inline.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Inline.get_inverse_event_shape(output_shape)` {#Inline.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.inverse(y, name='inverse', **condition_kwargs)` {#Inline.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -1504,6 +1988,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Inline.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Inline.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -1655,6 +2159,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.forward_event_shape(input_shape, name='forward_event_shape')` {#Invert.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Invert.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -1681,6 +2205,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.get_forward_event_shape(input_shape)` {#Invert.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Invert.get_inverse_event_shape(output_shape)` {#Invert.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.inverse(y, name='inverse', **condition_kwargs)` {#Invert.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -1738,6 +2304,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Invert.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Invert.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -1924,6 +2510,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.forward_event_shape(input_shape, name='forward_event_shape')` {#ScaleAndShift.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#ScaleAndShift.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -1950,6 +2556,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.get_forward_event_shape(input_shape)` {#ScaleAndShift.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.ScaleAndShift.get_inverse_event_shape(output_shape)` {#ScaleAndShift.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.inverse(y, name='inverse', **condition_kwargs)` {#ScaleAndShift.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -2007,6 +2655,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.inverse_event_shape(output_shape, name='inverse_event_shape')` {#ScaleAndShift.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#ScaleAndShift.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -2141,6 +2809,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.forward_event_shape(input_shape, name='forward_event_shape')` {#SigmoidCentered.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#SigmoidCentered.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -2167,6 +2855,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.get_forward_event_shape(input_shape)` {#SigmoidCentered.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.SigmoidCentered.get_inverse_event_shape(output_shape)` {#SigmoidCentered.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.inverse(y, name='inverse', **condition_kwargs)` {#SigmoidCentered.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -2224,6 +2954,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.inverse_event_shape(output_shape, name='inverse_event_shape')` {#SigmoidCentered.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#SigmoidCentered.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -2366,6 +3116,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.forward_event_shape(input_shape, name='forward_event_shape')` {#SoftmaxCentered.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#SoftmaxCentered.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -2392,6 +3162,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.get_forward_event_shape(input_shape)` {#SoftmaxCentered.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.get_inverse_event_shape(output_shape)` {#SoftmaxCentered.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse(y, name='inverse', **condition_kwargs)` {#SoftmaxCentered.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -2449,6 +3261,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse_event_shape(output_shape, name='inverse_event_shape')` {#SoftmaxCentered.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#SoftmaxCentered.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
@@ -2588,6 +3420,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.forward_event_shape(input_shape, name='forward_event_shape')` {#Softplus.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Softplus.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -2614,6 +3466,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.get_forward_event_shape(input_shape)` {#Softplus.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Softplus.get_inverse_event_shape(output_shape)` {#Softplus.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.inverse(y, name='inverse', **condition_kwargs)` {#Softplus.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -2671,6 +3565,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Softplus.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Softplus.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.CholeskyOuterProduct.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.CholeskyOuterProduct.md
index 5805851802..d2e81eafd0 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.CholeskyOuterProduct.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.CholeskyOuterProduct.md
@@ -69,6 +69,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.forward_event_shape(input_shape, name='forward_event_shape')` {#CholeskyOuterProduct.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#CholeskyOuterProduct.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -95,6 +115,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.get_forward_event_shape(input_shape)` {#CholeskyOuterProduct.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.get_inverse_event_shape(output_shape)` {#CholeskyOuterProduct.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse(y, name='inverse', **condition_kwargs)` {#CholeskyOuterProduct.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -152,6 +214,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse_event_shape(output_shape, name='inverse_event_shape')` {#CholeskyOuterProduct.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.CholeskyOuterProduct.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#CholeskyOuterProduct.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.SigmoidCentered.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.SigmoidCentered.md
index b7e279d4ee..aa32a7640a 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.SigmoidCentered.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard0/tf.contrib.distributions.bijector.SigmoidCentered.md
@@ -44,6 +44,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.forward_event_shape(input_shape, name='forward_event_shape')` {#SigmoidCentered.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#SigmoidCentered.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -70,6 +90,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.get_forward_event_shape(input_shape)` {#SigmoidCentered.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.SigmoidCentered.get_inverse_event_shape(output_shape)` {#SigmoidCentered.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.inverse(y, name='inverse', **condition_kwargs)` {#SigmoidCentered.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -127,6 +189,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SigmoidCentered.inverse_event_shape(output_shape, name='inverse_event_shape')` {#SigmoidCentered.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SigmoidCentered.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#SigmoidCentered.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard2/tf.contrib.distributions.bijector.Bijector.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard2/tf.contrib.distributions.bijector.Bijector.md
index b1f349e759..c59091bdd5 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard2/tf.contrib.distributions.bijector.Bijector.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard2/tf.contrib.distributions.bijector.Bijector.md
@@ -137,6 +137,10 @@ Subclass Requirements:
   functions to avoid computing the `inverse_log_det_jacobian` or the
   `inverse`, respectively.
 
+- Subclasses should implement `_get_forward_event_shape`,
+  `_forward_event_shape` (and `inverse` counterparts) if the transformation is
+  shape-changing.  By default the event-shape is assumed unchanged from input.
+
 Tips for implementing `_inverse` and `_inverse_log_det_jacobian`:
 
 - As case 3 [above] indicates, under some circumstances the inverse function
@@ -254,6 +258,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.forward_event_shape(input_shape, name='forward_event_shape')` {#Bijector.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Bijector.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -280,6 +304,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.get_forward_event_shape(input_shape)` {#Bijector.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Bijector.get_inverse_event_shape(output_shape)` {#Bijector.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.inverse(y, name='inverse', **condition_kwargs)` {#Bijector.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -337,6 +403,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Bijector.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Bijector.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Bijector.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Bijector.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard3/tf.contrib.distributions.bijector.ScaleAndShift.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard3/tf.contrib.distributions.bijector.ScaleAndShift.md
index bf1966d471..b0aed9abe8 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard3/tf.contrib.distributions.bijector.ScaleAndShift.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard3/tf.contrib.distributions.bijector.ScaleAndShift.md
@@ -110,6 +110,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.forward_event_shape(input_shape, name='forward_event_shape')` {#ScaleAndShift.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#ScaleAndShift.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -136,6 +156,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.get_forward_event_shape(input_shape)` {#ScaleAndShift.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.ScaleAndShift.get_inverse_event_shape(output_shape)` {#ScaleAndShift.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.inverse(y, name='inverse', **condition_kwargs)` {#ScaleAndShift.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -193,6 +255,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.ScaleAndShift.inverse_event_shape(output_shape, name='inverse_event_shape')` {#ScaleAndShift.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.ScaleAndShift.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#ScaleAndShift.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Invert.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Invert.md
index 41ced3f475..a2ebef5313 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Invert.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Invert.md
@@ -75,6 +75,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.forward_event_shape(input_shape, name='forward_event_shape')` {#Invert.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Invert.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -101,6 +121,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.get_forward_event_shape(input_shape)` {#Invert.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Invert.get_inverse_event_shape(output_shape)` {#Invert.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.inverse(y, name='inverse', **condition_kwargs)` {#Invert.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -158,6 +220,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Invert.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Invert.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Invert.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Invert.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Softplus.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Softplus.md
index dc85ab10ef..6204124a24 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Softplus.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard4/tf.contrib.distributions.bijector.Softplus.md
@@ -63,6 +63,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.forward_event_shape(input_shape, name='forward_event_shape')` {#Softplus.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Softplus.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -89,6 +109,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.get_forward_event_shape(input_shape)` {#Softplus.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Softplus.get_inverse_event_shape(output_shape)` {#Softplus.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.inverse(y, name='inverse', **condition_kwargs)` {#Softplus.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -146,6 +208,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Softplus.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Softplus.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Softplus.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Softplus.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Chain.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Chain.md
index a129c3edef..5045ed30c1 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Chain.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Chain.md
@@ -92,6 +92,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.forward_event_shape(input_shape, name='forward_event_shape')` {#Chain.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Chain.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -118,6 +138,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.get_forward_event_shape(input_shape)` {#Chain.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Chain.get_inverse_event_shape(output_shape)` {#Chain.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.inverse(y, name='inverse', **condition_kwargs)` {#Chain.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -175,6 +237,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Chain.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Chain.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Chain.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Chain.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Exp.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Exp.md
index 84eb7e4127..5e4631a940 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Exp.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard5/tf.contrib.distributions.bijector.Exp.md
@@ -66,6 +66,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.forward_event_shape(input_shape, name='forward_event_shape')` {#Exp.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Exp.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -92,6 +112,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.get_forward_event_shape(input_shape)` {#Exp.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Exp.get_inverse_event_shape(output_shape)` {#Exp.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.inverse(y, name='inverse', **condition_kwargs)` {#Exp.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -149,6 +211,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Exp.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Exp.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Exp.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Exp.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard6/tf.contrib.distributions.bijector.Identity.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard6/tf.contrib.distributions.bijector.Identity.md
index 29099fc926..0804a3f7bb 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard6/tf.contrib.distributions.bijector.Identity.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard6/tf.contrib.distributions.bijector.Identity.md
@@ -51,6 +51,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.forward_event_shape(input_shape, name='forward_event_shape')` {#Identity.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Identity.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -77,6 +97,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.get_forward_event_shape(input_shape)` {#Identity.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Identity.get_inverse_event_shape(output_shape)` {#Identity.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.inverse(y, name='inverse', **condition_kwargs)` {#Identity.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -134,6 +196,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Identity.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Identity.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Identity.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Identity.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard7/tf.contrib.distributions.bijector.Inline.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard7/tf.contrib.distributions.bijector.Inline.md
index 0e59026427..646dacb9fc 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard7/tf.contrib.distributions.bijector.Inline.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard7/tf.contrib.distributions.bijector.Inline.md
@@ -14,7 +14,7 @@ exp = Inline(
 The above example is equivalent to the `Bijector` `Exp(event_ndims=1)`.
 - - -
 
-#### `tf.contrib.distributions.bijector.Inline.__init__(forward_fn=None, inverse_fn=None, inverse_log_det_jacobian_fn=None, forward_log_det_jacobian_fn=None, is_constant_jacobian=False, validate_args=False, name='inline')` {#Inline.__init__}
+#### `tf.contrib.distributions.bijector.Inline.__init__(forward_fn=None, inverse_fn=None, inverse_log_det_jacobian_fn=None, forward_log_det_jacobian_fn=None, get_forward_event_shape_fn=None, forward_event_shape_fn=None, get_inverse_event_shape_fn=None, inverse_event_shape_fn=None, is_constant_jacobian=False, validate_args=False, name='inline')` {#Inline.__init__}
 
 Creates a `Bijector` from callables.
 
@@ -27,6 +27,14 @@ Creates a `Bijector` from callables.
     log o det o jacobian of the inverse transformation.
 *  <b>`forward_log_det_jacobian_fn`</b>: Python callable implementing the
     log o det o jacobian of the forward transformation.
+*  <b>`get_forward_event_shape_fn`</b>: Python callable implementing non-identical
+    static event shape changes. Default: shape is assumed unchanged.
+*  <b>`forward_event_shape_fn`</b>: Python callable implementing non-identical event
+    shape changes. Default: shape is assumed unchanged.
+*  <b>`get_inverse_event_shape_fn`</b>: Python callable implementing non-identical
+    static event shape changes. Default: shape is assumed unchanged.
+*  <b>`inverse_event_shape_fn`</b>: Python callable implementing non-identical event
+    shape changes. Default: shape is assumed unchanged.
 *  <b>`is_constant_jacobian`</b>: `Boolean` indicating that the Jacobian is constant
     for all input arguments.
 *  <b>`validate_args`</b>: `Boolean` indicating whether arguments should be checked
@@ -68,6 +76,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.forward_event_shape(input_shape, name='forward_event_shape')` {#Inline.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#Inline.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -94,6 +122,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.get_forward_event_shape(input_shape)` {#Inline.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.Inline.get_inverse_event_shape(output_shape)` {#Inline.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.inverse(y, name='inverse', **condition_kwargs)` {#Inline.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -151,6 +221,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.Inline.inverse_event_shape(output_shape, name='inverse_event_shape')` {#Inline.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.Inline.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#Inline.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.contrib.distributions.bijector.SoftmaxCentered.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.contrib.distributions.bijector.SoftmaxCentered.md
index 2601024ce3..580afec6c0 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.contrib.distributions.bijector.SoftmaxCentered.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.contrib.distributions.bijector.SoftmaxCentered.md
@@ -66,6 +66,26 @@ Returns the forward `Bijector` evaluation, i.e., X = g(Y).
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.forward_event_shape(input_shape, name='forward_event_shape')` {#SoftmaxCentered.forward_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `forward` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `forward`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.forward_log_det_jacobian(x, name='forward_log_det_jacobian', **condition_kwargs)` {#SoftmaxCentered.forward_log_det_jacobian}
 
 Returns both the forward_log_det_jacobian.
@@ -92,6 +112,48 @@ Returns both the forward_log_det_jacobian.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.get_forward_event_shape(input_shape)` {#SoftmaxCentered.get_forward_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `forward_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`input_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `forward` function.
+
+##### Returns:
+
+
+*  <b>`forward_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `forward`. Possibly unknown.
+
+
+- - -
+
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.get_inverse_event_shape(output_shape)` {#SoftmaxCentered.get_inverse_event_shape}
+
+Shape of a single sample from a single batch as a `TensorShape`.
+
+Same meaning as `inverse_event_shape`. May be only partially defined.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `TensorShape` indicating event-portion shape passed into
+    `inverse` function.
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `TensorShape` indicating event-portion shape after
+    applying `inverse`. Possibly unknown.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse(y, name='inverse', **condition_kwargs)` {#SoftmaxCentered.inverse}
 
 Returns the inverse `Bijector` evaluation, i.e., X = g^{-1}(Y).
@@ -149,6 +211,26 @@ See `inverse()`, `inverse_log_det_jacobian()` for more details.
 
 - - -
 
+#### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse_event_shape(output_shape, name='inverse_event_shape')` {#SoftmaxCentered.inverse_event_shape}
+
+Shape of a single sample from a single batch as an `int32` 1D `Tensor`.
+
+##### Args:
+
+
+*  <b>`output_shape`</b>: `Tensor`, `int32` vector indicating event-portion shape
+    passed into `inverse` function.
+*  <b>`name`</b>: name to give to the op
+
+##### Returns:
+
+
+*  <b>`inverse_event_shape`</b>: `Tensor`, `int32` vector indicating event-portion
+    shape after applying `inverse`.
+
+
+- - -
+
 #### `tf.contrib.distributions.bijector.SoftmaxCentered.inverse_log_det_jacobian(y, name='inverse_log_det_jacobian', **condition_kwargs)` {#SoftmaxCentered.inverse_log_det_jacobian}
 
 Returns the (log o det o Jacobian o inverse)(y).