Update generated Python Op docs.

Change: 136558601

6 files changed, 174 insertions, 58 deletions

diff --git a/tensorflow/g3doc/api_docs/python/contrib.distributions.md b/tensorflow/g3doc/api_docs/python/contrib.distributions.md
index f0e5da3d98..34ae9e2d94 100644
--- a/tensorflow/g3doc/api_docs/python/contrib.distributions.md
+++ b/tensorflow/g3doc/api_docs/python/contrib.distributions.md
@@ -20217,8 +20217,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20335,8 +20335,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20419,8 +20419,8 @@ Implements `(log o p o g^{-1})(y) + (log o det o J o g^{-1})(y)`,
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20458,8 +20458,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20611,8 +20611,8 @@ Implements `p(g^{-1}(y)) det|J(g^{-1}(y))|`, where `g^{-1}` is the
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20665,8 +20665,8 @@ Samples from the base distribution and then passes through
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -20714,8 +20714,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.distributions.TransformedDistribution.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.distributions.TransformedDistribution.md
index fc76841477..25a8707b38 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.distributions.TransformedDistribution.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.contrib.distributions.TransformedDistribution.md
@@ -182,8 +182,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -300,8 +300,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -384,8 +384,8 @@ Implements `(log o p o g^{-1})(y) + (log o det o J o g^{-1})(y)`,
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -423,8 +423,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -576,8 +576,8 @@ Implements `p(g^{-1}(y)) det|J(g^{-1}(y))|`, where `g^{-1}` is the
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -630,8 +630,8 @@ Samples from the base distribution and then passes through
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
@@ -679,8 +679,8 @@ Additional documentation from `TransformedDistribution`:
 
 ##### <b>`condition_kwargs`</b>:
 
-*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 *  <b>`bijector_kwargs`</b>: Python dictionary of arg names/values forwarded to the bijector.
+*  <b>`distribution_kwargs`</b>: Python dictionary of arg names/values forwarded to the distribution.
 
 ##### Args:
 
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.nn.conv1d.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.nn.conv1d.md
index f4c93a4328..d073ce7fb2 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.nn.conv1d.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard1/tf.nn.conv1d.md
@@ -2,19 +2,27 @@
 
 Computes a 1-D convolution given 3-D input and filter tensors.
 
-Given an input tensor of shape [batch, in_width, in_channels]
+Given an input tensor of shape
+  [batch, in_width, in_channels]
+if data_format is "NHWC", or
+  [batch, in_channels, in_width]
+if data_format is "NCHW",
 and a filter / kernel tensor of shape
 [filter_width, in_channels, out_channels], this op reshapes
 the arguments to pass them to conv2d to perform the equivalent
 convolution operation.
 
-Internally, this op reshapes the input tensors and invokes
-`tf.nn.conv2d`.  A tensor of shape [batch, in_width, in_channels]
-is reshaped to [batch, 1, in_width, in_channels], and the filter
-is reshaped to [1, filter_width, in_channels, out_channels].
-The result is then reshaped back to [batch, out_width, out_channels]
-(where out_width is a function of the stride and padding as in
-conv2d) and returned to the caller.
+Internally, this op reshapes the input tensors and invokes `tf.nn.conv2d`.
+For example, if `data_format` does not start with "NC", a tensor of shape
+  [batch, in_width, in_channels]
+is reshaped to
+  [batch, 1, in_width, in_channels],
+and the filter is reshaped to
+  [1, filter_width, in_channels, out_channels].
+The result is then reshaped back to
+  [batch, out_width, out_channels]
+(where out_width is a function of the stride and padding as in conv2d) and
+returned to the caller.
 
 ##### Args:
 
@@ -35,3 +43,8 @@ conv2d) and returned to the caller.
 
   A `Tensor`.  Has the same type as input.
 
+##### Raises:
+
+
+*  <b>`ValueError`</b>: if `data_format` is invalid.
+
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.convolution.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.convolution.md
index cff3950ed6..f1ed0e2f53 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.convolution.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.convolution.md
@@ -1,4 +1,4 @@
-### `tf.nn.convolution(input, filter, padding, strides=None, dilation_rate=None, name=None)` {#convolution}
+### `tf.nn.convolution(input, filter, padding, strides=None, dilation_rate=None, name=None, data_format=None)` {#convolution}
 
 Computes sums of N-D convolutions (actually cross-correlation).
 
@@ -8,7 +8,8 @@ convolution, based on the French word "trous" meaning holes in English) via
 the optional `dilation_rate` parameter.  Currently, however, output striding
 is not supported for atrous convolutions.
 
-Specifically, given rank (N+2) `input` Tensor of shape
+Specifically, in the case that `data_format` does not start with "NC", given
+a rank (N+2) `input` Tensor of shape
 
   [num_batches,
    input_spatial_shape[0],
@@ -33,25 +34,36 @@ position (x[0], ..., x[N-1]):
 
       sum_{z[0], ..., z[N-1], q}
 
-          filters[z[0], ..., z[N-1], q, k] *
+          filter[z[0], ..., z[N-1], q, k] *
           padded_input[b,
                        x[0]*strides[0] + dilation_rate[0]*z[0],
                        ...,
                        x[N-1]*strides[N-1] + dilation_rate[N-1]*z[N-1],
-                       q],
+                       q]
 
 where `padded_input` is obtained by zero padding the input using an effective
 spatial filter shape of `(spatial_filter_shape-1) * dilation_rate + 1` and
 output striding `strides` as described in the
 [comment here](https://www.tensorflow.org/api_docs/python/nn.html#convolution).
 
+In the case that `data_format` does start with `"NC"`, the `input` and output
+(but not the `filter`) are simply transposed as follows:
+
+  convolution(input, data_format, **kwargs) =
+    tf.transpose(convolution(tf.transpose(input, [0] + range(2,N+2) + [1]),
+                             **kwargs),
+                 [0, N+1] + range(1, N+1))
+
 It is required that 1 <= N <= 3.
 
 ##### Args:
 
 
 *  <b>`input`</b>: An N-D `Tensor` of type `T`, of shape
-    `[batch_size] + input_spatial_shape + [in_channels]`.
+    `[batch_size] + input_spatial_shape + [in_channels]` if data_format does
+    not start with "NC" (default), or
+    `[batch_size, in_channels] + input_spatial_shape` if data_format starts
+    with "NC".
 *  <b>`filter`</b>: An N-D `Tensor` with the same type as `input` and shape
     `spatial_filter_shape + [in_channels, out_channels]`.
 *  <b>`padding`</b>: A string, either `"VALID"` or `"SAME"`. The padding algorithm.
@@ -67,13 +79,24 @@ It is required that 1 <= N <= 3.
     filter in each spatial dimension i.  If any value of dilation_rate is > 1,
     then all values of strides must be 1.
 *  <b>`name`</b>: Optional name for the returned tensor.
+*  <b>`data_format`</b>: A string or None.  Specifies whether the channel dimension of
+    the `input` and output is the last dimension (default, or if `data_format`
+    does not start with "NC"), or the second dimension (if `data_format`
+    starts with "NC").  For N=1, the valid values are "NWC" (default) and
+    "NCW".  For N=2, the valid values are "NHWC" (default) and "NCHW".  For
+    N=3, the valid value is "NDHWC".
 
 ##### Returns:
 
-  A `Tensor` with the same type as `value` of shape
+  A `Tensor` with the same type as `input` of shape
+
+      `[batch_size] + output_spatial_shape + [out_channels]`
+
+  if data_format is None or does not start with "NC", or
 
-      `[batch_size] + output_spatial_shape + [out_channels]`,
+      `[batch_size, out_channels] + output_spatial_shape`
 
+  if data_format starts with "NC",
   where `output_spatial_shape` depends on the value of `padding`.
 
   If padding == "SAME":
@@ -88,6 +111,6 @@ It is required that 1 <= N <= 3.
 ##### Raises:
 
 
-*  <b>`ValueError`</b>: If input/output depth does not match `filter` shape, or if
-    padding is other than `"VALID"` or `"SAME"`.
+*  <b>`ValueError`</b>: If input/output depth does not match `filter` shape, if padding
+    is other than `"VALID"` or `"SAME"`, or if data_format is invalid.
 
diff --git a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.pool.md b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.pool.md
index b97f4312ee..98a70fde53 100644
--- a/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.pool.md
+++ b/tensorflow/g3doc/api_docs/python/functions_and_classes/shard8/tf.nn.pool.md
@@ -1,8 +1,8 @@
-### `tf.nn.pool(input, window_shape, pooling_type, padding, dilation_rate=None, strides=None, name=None)` {#pool}
+### `tf.nn.pool(input, window_shape, pooling_type, padding, dilation_rate=None, strides=None, name=None, data_format=None)` {#pool}
 
 Performs an N-D pooling operation.
 
-Computes for
+In the case that `data_format` does not start with "NC", computes for
     0 <= b < batch_size,
     0 <= x[i] < output_spatial_shape[i],
     0 <= c < num_channels:
@@ -20,12 +20,22 @@ and pad_before is defined based on the value of `padding` as described in the
 [comment here](https://www.tensorflow.org/api_docs/python/nn.html#convolution).
 The reduction never includes out-of-bounds positions.
 
+In the case that `data_format` starts with `"NC"`, the `input` and output are
+simply transposed as follows:
+
+  pool(input, data_format, **kwargs) =
+    tf.transpose(pool(tf.transpose(input, [0] + range(2,N+2) + [1]),
+                      **kwargs),
+                 [0, N+1] + range(1, N+1))
+
 ##### Args:
 
 
 *  <b>`input`</b>: Tensor of rank N+2, of shape
-    [batch_size] + input_spatial_shape + [num_channels].
-    Pooling happens over the spatial dimensions only.
+    `[batch_size] + input_spatial_shape + [num_channels]` if data_format does
+    not start with "NC" (default), or
+    `[batch_size, num_channels] + input_spatial_shape` if data_format starts
+    with "NC".  Pooling happens over the spatial dimensions only.
 *  <b>`window_shape`</b>: Sequence of N ints >= 1.
 *  <b>`pooling_type`</b>: Specifies pooling operation, must be "AVG" or "MAX".
 *  <b>`padding`</b>: The padding algorithm, must be "SAME" or "VALID".
@@ -37,11 +47,23 @@ The reduction never includes out-of-bounds positions.
     If any value of strides is > 1, then all values of dilation_rate must be
     1.
 *  <b>`name`</b>: Optional. Name of the op.
+*  <b>`data_format`</b>: A string or None.  Specifies whether the channel dimension of
+    the `input` and output is the last dimension (default, or if `data_format`
+    does not start with "NC"), or the second dimension (if `data_format`
+    starts with "NC").  For N=1, the valid values are "NWC" (default) and
+    "NCW".  For N=2, the valid values are "NHWC" (default) and "NCHW".  For
+    N=3, the valid value is "NDHWC".
 
 ##### Returns:
 
   Tensor of rank N+2, of shape
-    [batch_size] + output_spatial_shape + [num_channels],
+    [batch_size] + output_spatial_shape + [num_channels]
+
+  if data_format is None or does not start with "NC", or
+
+    [batch_size, num_channels] + output_spatial_shape
+
+  if data_format starts with "NC",
   where `output_spatial_shape` depends on the value of padding:
 
   If padding = "SAME":
diff --git a/tensorflow/g3doc/api_docs/python/nn.md b/tensorflow/g3doc/api_docs/python/nn.md
index 7ce8700e36..6479d5034e 100644
--- a/tensorflow/g3doc/api_docs/python/nn.md
+++ b/tensorflow/g3doc/api_docs/python/nn.md
@@ -309,7 +309,7 @@ concatenated.
 
 - - -
 
-### `tf.nn.convolution(input, filter, padding, strides=None, dilation_rate=None, name=None)` {#convolution}
+### `tf.nn.convolution(input, filter, padding, strides=None, dilation_rate=None, name=None, data_format=None)` {#convolution}
 
 Computes sums of N-D convolutions (actually cross-correlation).
 
@@ -319,7 +319,8 @@ convolution, based on the French word "trous" meaning holes in English) via
 the optional `dilation_rate` parameter.  Currently, however, output striding
 is not supported for atrous convolutions.
 
-Specifically, given rank (N+2) `input` Tensor of shape
+Specifically, in the case that `data_format` does not start with "NC", given
+a rank (N+2) `input` Tensor of shape
 
   [num_batches,
    input_spatial_shape[0],
@@ -344,25 +345,36 @@ position (x[0], ..., x[N-1]):
 
       sum_{z[0], ..., z[N-1], q}
 
-          filters[z[0], ..., z[N-1], q, k] *
+          filter[z[0], ..., z[N-1], q, k] *
           padded_input[b,
                        x[0]*strides[0] + dilation_rate[0]*z[0],
                        ...,
                        x[N-1]*strides[N-1] + dilation_rate[N-1]*z[N-1],
-                       q],
+                       q]
 
 where `padded_input` is obtained by zero padding the input using an effective
 spatial filter shape of `(spatial_filter_shape-1) * dilation_rate + 1` and
 output striding `strides` as described in the
 [comment here](https://www.tensorflow.org/api_docs/python/nn.html#convolution).
 
+In the case that `data_format` does start with `"NC"`, the `input` and output
+(but not the `filter`) are simply transposed as follows:
+
+  convolution(input, data_format, **kwargs) =
+    tf.transpose(convolution(tf.transpose(input, [0] + range(2,N+2) + [1]),
+                             **kwargs),
+                 [0, N+1] + range(1, N+1))
+
 It is required that 1 <= N <= 3.
 
 ##### Args:
 
 
 *  <b>`input`</b>: An N-D `Tensor` of type `T`, of shape
-    `[batch_size] + input_spatial_shape + [in_channels]`.
+    `[batch_size] + input_spatial_shape + [in_channels]` if data_format does
+    not start with "NC" (default), or
+    `[batch_size, in_channels] + input_spatial_shape` if data_format starts
+    with "NC".
 *  <b>`filter`</b>: An N-D `Tensor` with the same type as `input` and shape
     `spatial_filter_shape + [in_channels, out_channels]`.
 *  <b>`padding`</b>: A string, either `"VALID"` or `"SAME"`. The padding algorithm.
@@ -378,13 +390,24 @@ It is required that 1 <= N <= 3.
     filter in each spatial dimension i.  If any value of dilation_rate is > 1,
     then all values of strides must be 1.
 *  <b>`name`</b>: Optional name for the returned tensor.
+*  <b>`data_format`</b>: A string or None.  Specifies whether the channel dimension of
+    the `input` and output is the last dimension (default, or if `data_format`
+    does not start with "NC"), or the second dimension (if `data_format`
+    starts with "NC").  For N=1, the valid values are "NWC" (default) and
+    "NCW".  For N=2, the valid values are "NHWC" (default) and "NCHW".  For
+    N=3, the valid value is "NDHWC".
 
 ##### Returns:
 
-  A `Tensor` with the same type as `value` of shape
+  A `Tensor` with the same type as `input` of shape
+
+      `[batch_size] + output_spatial_shape + [out_channels]`
+
+  if data_format is None or does not start with "NC", or
 
-      `[batch_size] + output_spatial_shape + [out_channels]`,
+      `[batch_size, out_channels] + output_spatial_shape`
 
+  if data_format starts with "NC",
   where `output_spatial_shape` depends on the value of `padding`.
 
   If padding == "SAME":
@@ -399,8 +422,8 @@ It is required that 1 <= N <= 3.
 ##### Raises:
 
 
-*  <b>`ValueError`</b>: If input/output depth does not match `filter` shape, or if
-    padding is other than `"VALID"` or `"SAME"`.
+*  <b>`ValueError`</b>: If input/output depth does not match `filter` shape, if padding
+    is other than `"VALID"` or `"SAME"`, or if data_format is invalid.
 
 
 - - -
@@ -710,19 +733,27 @@ deconvolution.
 
 Computes a 1-D convolution given 3-D input and filter tensors.
 
-Given an input tensor of shape [batch, in_width, in_channels]
+Given an input tensor of shape
+  [batch, in_width, in_channels]
+if data_format is "NHWC", or
+  [batch, in_channels, in_width]
+if data_format is "NCHW",
 and a filter / kernel tensor of shape
 [filter_width, in_channels, out_channels], this op reshapes
 the arguments to pass them to conv2d to perform the equivalent
 convolution operation.
 
-Internally, this op reshapes the input tensors and invokes
-`tf.nn.conv2d`.  A tensor of shape [batch, in_width, in_channels]
-is reshaped to [batch, 1, in_width, in_channels], and the filter
-is reshaped to [1, filter_width, in_channels, out_channels].
-The result is then reshaped back to [batch, out_width, out_channels]
-(where out_width is a function of the stride and padding as in
-conv2d) and returned to the caller.
+Internally, this op reshapes the input tensors and invokes `tf.nn.conv2d`.
+For example, if `data_format` does not start with "NC", a tensor of shape
+  [batch, in_width, in_channels]
+is reshaped to
+  [batch, 1, in_width, in_channels],
+and the filter is reshaped to
+  [1, filter_width, in_channels, out_channels].
+The result is then reshaped back to
+  [batch, out_width, out_channels]
+(where out_width is a function of the stride and padding as in conv2d) and
+returned to the caller.
 
 ##### Args:
 
@@ -743,6 +774,11 @@ conv2d) and returned to the caller.
 
   A `Tensor`.  Has the same type as input.
 
+##### Raises:
+
+
+*  <b>`ValueError`</b>: if `data_format` is invalid.
+
 
 - - -
 
@@ -1116,11 +1152,11 @@ For more details on fractional max pooling, see this paper:
 
 - - -
 
-### `tf.nn.pool(input, window_shape, pooling_type, padding, dilation_rate=None, strides=None, name=None)` {#pool}
+### `tf.nn.pool(input, window_shape, pooling_type, padding, dilation_rate=None, strides=None, name=None, data_format=None)` {#pool}
 
 Performs an N-D pooling operation.
 
-Computes for
+In the case that `data_format` does not start with "NC", computes for
     0 <= b < batch_size,
     0 <= x[i] < output_spatial_shape[i],
     0 <= c < num_channels:
@@ -1138,12 +1174,22 @@ and pad_before is defined based on the value of `padding` as described in the
 [comment here](https://www.tensorflow.org/api_docs/python/nn.html#convolution).
 The reduction never includes out-of-bounds positions.
 
+In the case that `data_format` starts with `"NC"`, the `input` and output are
+simply transposed as follows:
+
+  pool(input, data_format, **kwargs) =
+    tf.transpose(pool(tf.transpose(input, [0] + range(2,N+2) + [1]),
+                      **kwargs),
+                 [0, N+1] + range(1, N+1))
+
 ##### Args:
 
 
 *  <b>`input`</b>: Tensor of rank N+2, of shape
-    [batch_size] + input_spatial_shape + [num_channels].
-    Pooling happens over the spatial dimensions only.
+    `[batch_size] + input_spatial_shape + [num_channels]` if data_format does
+    not start with "NC" (default), or
+    `[batch_size, num_channels] + input_spatial_shape` if data_format starts
+    with "NC".  Pooling happens over the spatial dimensions only.
 *  <b>`window_shape`</b>: Sequence of N ints >= 1.
 *  <b>`pooling_type`</b>: Specifies pooling operation, must be "AVG" or "MAX".
 *  <b>`padding`</b>: The padding algorithm, must be "SAME" or "VALID".
@@ -1155,11 +1201,23 @@ The reduction never includes out-of-bounds positions.
     If any value of strides is > 1, then all values of dilation_rate must be
     1.
 *  <b>`name`</b>: Optional. Name of the op.
+*  <b>`data_format`</b>: A string or None.  Specifies whether the channel dimension of
+    the `input` and output is the last dimension (default, or if `data_format`
+    does not start with "NC"), or the second dimension (if `data_format`
+    starts with "NC").  For N=1, the valid values are "NWC" (default) and
+    "NCW".  For N=2, the valid values are "NHWC" (default) and "NCHW".  For
+    N=3, the valid value is "NDHWC".
 
 ##### Returns:
 
   Tensor of rank N+2, of shape
-    [batch_size] + output_spatial_shape + [num_channels],
+    [batch_size] + output_spatial_shape + [num_channels]
+
+  if data_format is None or does not start with "NC", or
+
+    [batch_size, num_channels] + output_spatial_shape
+
+  if data_format starts with "NC",
   where `output_spatial_shape` depends on the value of padding:
 
   If padding = "SAME":