Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2298
Star 20931
Fork 5422
P
Paddle
提交 66ea7184 编写于 作者: H haowang101779990
浏览文件
操作

en api improve format Dec 27 

test=develop
上级 988bc2b5
显示空白变更内容
内联 并排
Showing with 379 addition and 291 deletion
python/paddle/fluid/data_feeder.py
浏览文件 @ 66ea7184
......@@ -272,8 +272,7 @@ class DataFeeder(object):
dict: the result of conversion.
Raises:
ValueError: If drop_last is False and the data batch which cannot
fit for devices.
ValueError: If drop_last is False and the data batch which cannot fit for devices.
"""
def __reader_creator__():
......
python/paddle/fluid/framework.py
浏览文件 @ 66ea7184
......@@ -1646,8 +1646,8 @@ class Program(object):
parameters, e.g., :code:`trainable`, :code:`optimize_attr`, need
to print.
Returns
(str): The debug string.
Returns:
str : The debug string.
Raises:
ValueError: If any of required fields is not set and throw_on_error is
......
python/paddle/fluid/layers/control_flow.py
浏览文件 @ 66ea7184
......@@ -1452,6 +1452,7 @@ class DynamicRNN(object):
def step_input(self, x):
"""
Mark a sequence as a dynamic RNN input.
Args:
x(Variable): The input sequence.
......@@ -1505,6 +1506,7 @@ class DynamicRNN(object):
"""
Mark a variable as a RNN input. The input will not be scattered into
time steps.
Args:
x(Variable): The input variable.
......@@ -1629,13 +1631,11 @@ class DynamicRNN(object):
Args:
init(Variable|None): The initialized variable.
shape(list|tuple): The memory shape. NOTE the shape does not contain
batch_size.
shape(list|tuple): The memory shape. NOTE the shape does not contain batch_size.
value(float): the initalized value.
need_reorder(bool): True if the initialized memory depends on the
input sample.
need_reorder(bool): True if the initialized memory depends on the input sample.
dtype(str|numpy.dtype): The data type of the initialized memory.
......@@ -1714,6 +1714,7 @@ class DynamicRNN(object):
"""
Update the memory from ex_mem to new_mem. NOTE that the shape and data
type of :code:`ex_mem` and :code:`new_mem` must be same.
Args:
ex_mem(Variable): the memory variable.
new_mem(Variable): the plain variable generated in RNN block.
......
python/paddle/fluid/layers/detection.py
浏览文件 @ 66ea7184
......@@ -65,7 +65,7 @@ def rpn_target_assign(bbox_pred,
rpn_negative_overlap=0.3,
use_random=True):
"""
** Target Assign Layer for region proposal network (RPN) in Faster-RCNN detection. **
**Target Assign Layer for region proposal network (RPN) in Faster-RCNN detection.**
This layer can be, for given the Intersection-over-Union (IoU) overlap
between anchors and ground truth boxes, to assign classification and
......@@ -148,6 +148,7 @@ def rpn_target_assign(bbox_pred,
cls_logits=cls_logits,
anchor_box=anchor_box,
gt_boxes=gt_boxes)
"""
helper = LayerHelper('rpn_target_assign', **locals())
......@@ -1525,20 +1526,23 @@ def anchor_generator(input,
anchors, e.g. [0.5, 1.0, 2.0].
variance(list|tuple): The variances to be used in box regression deltas.
Default:[0.1, 0.1, 0.2, 0.2].
stride(list|turple): The anchors stride across width and height,
e.g. [16.0, 16.0]
stride(list|turple): The anchors stride across width and height,e.g. [16.0, 16.0]
offset(float): Prior boxes center offset. Default: 0.5
name(str): Name of the prior box op. Default: None.
Returns:
Anchors(Variable): The output anchors with a layout of [H, W, num_anchors, 4].
H is the height of input, W is the width of input,
num_anchors is the box count of each position.
Anchors(Variable),Variances(Variable):
two variables:
- Anchors(Variable): The output anchors with a layout of [H, W, num_anchors, 4]. \
H is the height of input, W is the width of input, \
num_anchors is the box count of each position. \
Each anchor is in (xmin, ymin, xmax, ymax) format an unnormalized.
Variances(Variable): The expanded variances of anchors
with a layout of [H, W, num_priors, 4].
H is the height of input, W is the width of input
num_anchors is the box count of each position.
- Variances(Variable): The expanded variances of anchors \
with a layout of [H, W, num_priors, 4]. \
H is the height of input, W is the width of input \
num_anchors is the box count of each position. \
Each variance is in (xcenter, ycenter, w, h) format.
......@@ -1748,7 +1752,7 @@ def generate_proposals(scores,
eta=1.0,
name=None):
"""
** Generate proposal Faster-RCNN **
**Generate proposal Faster-RCNN**
This operation proposes RoIs according to each box with their probability to be a foreground object and
the box can be calculated by anchors. Bbox_deltais and scores to be an object are the output of RPN. Final proposals
......@@ -1762,7 +1766,6 @@ def generate_proposals(scores,
4. Remove predicted boxes with small area.
5. Apply NMS to get final proposals as output.
Args:
scores(Variable): A 4-D Tensor with shape [N, A, H, W] represents the probability for each box to be an object.
N is batch size, A is number of anchors, H and W are height and width of the feature map.
......@@ -1777,6 +1780,7 @@ def generate_proposals(scores,
nms_thresh(float): Threshold in NMS, 0.5 by default.
min_size(float): Remove predicted boxes with either height or width < min_size. 0.1 by default.
eta(float): Apply in adaptive NMS, if adaptive threshold > 0.5, adaptive_threshold = adaptive_threshold * eta in each iteration.
"""
helper = LayerHelper('generate_proposals', **locals())
......
python/paddle/fluid/layers/io.py
浏览文件 @ 66ea7184
......@@ -949,12 +949,11 @@ def shuffle(reader, buffer_size):
is determined by argument buf_size.
Args:
param reader: the original reader whose output will be shuffled.
type reader: callable
param buf_size: shuffle buffer size.
type buf_size: int
return: the new reader whose output is shuffled.
rtype: callable
reader(callable): the original reader whose output will be shuffled.
buf_size(int): shuffle buffer size.
Returns:
callable: the new reader whose output is shuffled.
"""
return __create_unshared_decorated_reader__(
'create_shuffle_reader', reader, {'buffer_size': int(buffer_size)})
......
python/paddle/fluid/layers/nn.py
浏览文件 @ 66ea7184
......@@ -233,7 +233,7 @@ def fc(input,
dimensions will be flatten to form the first dimension of the final matrix (height of
the matrix), and the rest `rank(X) - num_flatten_dims` dimensions are flattened to
form the second dimension of the final matrix (width of the matrix). For example, suppose
`X` is a 6-dimensional tensor with a shape [2, 3, 4, 5, 6], and `num_flatten_dims` = 3.
`X` is a 5-dimensional tensor with a shape [2, 3, 4, 5, 6], and `num_flatten_dims` = 3.
Then, the flattened matrix will have a shape [2 x 3 x 4, 5 x 6] = [24, 30].
param_attr (ParamAttr|list of ParamAttr, default None): The parameter attribute for learnable
parameters/weights of this layer.
......@@ -505,31 +505,33 @@ def lstm(input,
In the forward pass the output ht and cell output ct for a given iteration can be computed from the recurrent input ht-1,
the cell input ct-1 and the previous layer input xt given matrices W, R and biases bW, bR from the following equations:
$$ i_t = \\sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + bx_i + bh_i) $$
.. math::
i_t &= \sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + bx_i + bh_i)
$$ f_t = \\sigma(W_{fx}x_{t} + W_{fh}h_{t-1} + bx_f + bh_f) $$
f_t &= \sigma(W_{fx}x_{t} + W_{fh}h_{t-1} + bx_f + bh_f)
$$ o_t = \\sigma(W_{ox}x_{t} + W_{oh}h_{t-1} + bx_o + bh_o) $$
o_t &= \sigma(W_{ox}x_{t} + W_{oh}h_{t-1} + bx_o + bh_o)
$$ \\tilde{c_t} = tanh(W_{cx}x_t + W_{ch}h_{t-1} + bx_c + bh_c) $$
\\tilde{c_t} &= tanh(W_{cx}x_t + W_{ch}h_{t-1} + bx_c + bh_c)
$$ c_t = f_t \\odot c_{t-1} + i_t \\odot \\tilde{c_t} $$
c_t &= f_t \odot c_{t-1} + i_t \odot \\tilde{c_t}
$$ h_t = o_t \\odot tanh(c_t) $$
h_t &= o_t \odot tanh(c_t)
- W terms denote weight matrices (e.g. $W_{ix}$ is the matrix
- $W$ terms denote weight matrices (e.g. $W_{ix}$ is the matrix
of weights from the input gate to the input)
- The b terms denote bias vectors ($bx_i$ and $bh_i$ are the input gate bias vector).
- sigmoid is the logistic sigmoid function.
- $i, f, o$ and $c$ are the input gate, forget gate, output gate,
and cell activation vectors, respectively, all of which have the same size as
the cell output activation vector $h$.
- The $\odot$ is the element-wise product of the vectors.
- `tanh` is the activation functions.
- $\tilde{c_t}$ is also called candidate hidden state,
- The :math:`\odot` is the element-wise product of the vectors.
- :math:`tanh` is the activation functions.
- :math:`\\tilde{c_t}` is also called candidate hidden state,
which is computed based on the current input and the previous hidden state.
Where sigmoid is the sigmoid operator: sigmoid(x) = 1 / (1 + e^-x), * represents a point-wise multiplication,
Where sigmoid is the sigmoid operator: :math:`sigmoid(x) = 1 / (1 + e^{-x})` , * represents a point-wise multiplication,
X represensts a matrix multiplication
......@@ -556,13 +558,17 @@ def lstm(input,
Returns:
rnn_out(Tensor): result of LSTM hidden, shape is (seq_len x batch_size x hidden_size)
rnn_out(Tensor),last_h(Tensor),last_c(Tensor):
Three tensors, rnn_out, last_h, last_c:
- rnn_out is result of LSTM hidden, shape is (seq_len x batch_size x hidden_size) \
if is_bidirec set to True, shape will be ( seq_len x batch_sze x hidden_size*2)
last_h(Tensor): the hidden state of the last step of LSTM
shape is ( num_layers x batch_size x hidden_size )
- last_h is the hidden state of the last step of LSTM \
shape is ( num_layers x batch_size x hidden_size ) \
if is_bidirec set to True, shape will be ( num_layers*2 x batch_size x hidden_size)
last_c(Tensor): the cell state of the last step of LSTM
shape is ( num_layers x batch_size x hidden_size )
- last_c(Tensor): the cell state of the last step of LSTM \
shape is ( num_layers x batch_size x hidden_size ) \
if is_bidirec set to True, shape will be ( num_layers*2 x batch_size x hidden_size)
......@@ -1220,6 +1226,8 @@ def dropout(x,
probability) the outputs of some units to zero, while others are remain
unchanged.
dropout op can be removed from the program to make the program more efficient.
Args:
x (Variable): The input tensor variable.
dropout_prob (float): Probability of setting units to zero.
......@@ -1230,20 +1238,22 @@ def dropout(x,
units will be dropped. DO NOT use a fixed seed in training.
name (str|None): A name for this layer(optional). If set None, the layer
will be named automatically.
dropout_implementation(string): ['downgrade_in_infer'(defauld)|'upscale_in_train']
dropout_implementation(string): ['downgrade_in_infer'(default)|'upscale_in_train']
1. downgrade_in_infer(default), downgrade the outcome at inference
train: out = input * mask
inference: out = input * dropout_prob
(make is a tensor same shape with input, value is 0 or 1
- train: out = input * mask
- inference: out = input * dropout_prob
(mask is a tensor same shape with input, value is 0 or 1
ratio of 0 is dropout_prob)
2. upscale_in_train, upscale the outcome at training time
train: out = input * mask / ( 1.0 - dropout_prob )
inference: out = input
(make is a tensor same shape with input, value is 0 or 1
ratio of 0 is dropout_prob)
dropout op can be removed from the program.
the program will be efficient
- train: out = input * mask / ( 1.0 - dropout_prob )
- inference: out = input
(mask is a tensor same shape with input, value is 0 or 1
ratio of 0 is dropout_prob)
Returns:
......@@ -1333,11 +1343,15 @@ def cross_entropy(input, label, soft_label=False, ignore_index=kIgnoreIndex):
A 2-D tensor with shape [N x 1], the cross entropy loss.
Raises:
`ValueError`: 1) the 1st dimension of `input` and `label` are not equal.
2) when `soft_label == True`, and the 2nd dimension of
`input` and `label` are not equal.
3) when `soft_label == False`, and the 2nd dimension of
`label` is not 1.
ValueError:
1. the 1st dimension of ``input`` and ``label`` are not equal.
2. when ``soft_label == True``, and the 2nd dimension of
``input`` and ``label`` are not equal.
3. when ``soft_label == False``, and the 2nd dimension of
``label`` is not 1.
Examples:
.. code-block:: python
......@@ -1458,7 +1472,7 @@ def chunk_eval(input,
F1-score of chunk detection.
For some basics of chunking, please refer to
'Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>'.
`Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>`_ .
ChunkEvalOp computes the precision, recall, and F1-score of chunk detection,
and supports IOB, IOE, IOBES and IO (also known as plain) tagging schemes.
......@@ -2292,7 +2306,8 @@ def sequence_slice(input, offset, length, name=None):
out.lod = [[2, 1]],
out.dims = (3, 2).
NOTE: The first dimension size of **input**, **offset** and **length**
Note:
The first dimension size of **input**, **offset** and **length**
should be equal. The **offset** should start from 0.
Args:
......@@ -3013,7 +3028,7 @@ def group_norm(input,
"""
**Group Normalization Layer**
Refer to `Group Normalization <https://arxiv.org/abs/1803.08494>`
Refer to `Group Normalization <https://arxiv.org/abs/1803.08494>`_ .
Args:
input(Variable): The input tensor variable.
......@@ -3140,8 +3155,8 @@ def conv2d_transpose(input,
H^\prime_{out} &= (H_{in} - 1) * strides[0] - 2 * paddings[0] + dilations[0] * (H_f - 1) + 1 \\\\
W^\prime_{out} &= (W_{in} - 1) * strides[1] - 2 * paddings[1] + dilations[1] * (W_f - 1) + 1 \\\\
H_{out} \in [ H^\prime_{out}, H^\prime_{out} + strides[0] ) \\\\
W_{out} \in [ W^\prime_{out}, W^\prime_{out} + strides[1] )
H_{out} &\in [ H^\prime_{out}, H^\prime_{out} + strides[0] ) \\\\
W_{out} &\in [ W^\prime_{out}, W^\prime_{out} + strides[1] )
Args:
input(Variable): The input image with [N, C, H, W] format.
......@@ -4704,9 +4719,9 @@ def ctc_greedy_decoder(input, blank, name=None):
name (str): The name of this layer. It is optional.
Returns:
Variable: CTC greedy decode result which is a 2-D tensor with shape [Lp, 1].
'Lp' is the sum if all output sequences' length. If all the sequences
in result were empty, the result LoDTensor will be [-1] with
Variable: CTC greedy decode result which is a 2-D tensor with shape [Lp, 1]. \
'Lp' is the sum if all output sequences' length. If all the sequences \
in result were empty, the result LoDTensor will be [-1] with \
LoD [[]] and dims [1, 1].
Examples:
......@@ -5072,6 +5087,7 @@ def hsigmoid(input,
<http://www.iro.umontreal.ca/~lisa/pointeurs/hierarchical-nnlm-aistats05.pdf>`_
And if you want to use the costumed tree by set 'is_custom' as true you may need to do following things first:
1. using your word dict to build a binary tree, each leaf node should be an word of your word dict
2. build a dict to store word_id -> word's leaf to root path, we call it path_table.
3. build a dict to store word_id -> code of word's leaf to root path, we call it path_code. Code
......@@ -5079,7 +5095,6 @@ def hsigmoid(input,
4. now, each word should has its path and code along the path, you can pass a batch of path and code
related to the same batch of inputs.
Args:
input (Variable): The input tensor variable with shape
:math:`[N \\times D]`, where :math:`N` is the size of mini-batch,
......@@ -5485,11 +5500,11 @@ def softmax_with_cross_entropy(logits,
.. math::
max_j = \\max_{i=0}^{K}{\\text{logit}_i}
max_j &= \\max_{i=0}^{K}{\\text{logit}_i}
log\\_max\\_sum_j = \\log\\sum_{i=0}^{K}\\exp(logit_i - max_j)
log\\_max\\_sum_j &= \\log\\sum_{i=0}^{K}\\exp(logit_i - max_j)
softmax_j = \\exp(logit_j - max_j - {log\\_max\\_sum}_j)
softmax_j &= \\exp(logit_j - max_j - {log\\_max\\_sum}_j)
and then cross entropy loss is calculated by softmax and label.
......@@ -5515,10 +5530,10 @@ def softmax_with_cross_entropy(logits,
along with the cross entropy loss. Default: False
Returns:
Variable or Tuple of two Variables: Return the cross entropy loss if
`return_softmax` is False, otherwise the tuple
(loss, softmax), where the cross entropy loss is
a 2-D tensor with shape [N x 1], and softmax is a
Variable or Tuple of two Variables: Return the cross entropy loss if \
`return_softmax` is False, otherwise the tuple \
(loss, softmax), where the cross entropy loss is \
a 2-D tensor with shape [N x 1], and softmax is a \
2-D tensor with shape [N x K].
Examples:
......@@ -5792,15 +5807,21 @@ def squeeze(input, axes, name=None):
the single dimensions will be removed from the shape. If an axis is
selected with shape entry not equal to one, an error is raised.
Examples:
For example:
.. code-block:: text
Case 1:
Given
X.shape = (1, 3, 1, 5)
and
axes = [0]
we get:
Out.shape = (3, 1, 5)
Case 2:
Given
X.shape = (1, 3, 1, 5)
and
......@@ -5842,6 +5863,9 @@ def unsqueeze(input, axes, name=None):
Dimension indices in axes are as seen in the output tensor.
For example:
.. code-block:: text
Given a tensor such that tensor with shape [3, 4, 5],
then Unsqueezed tensor with axes=[0, 4] has shape [1, 3, 4, 5, 1].
......@@ -6729,8 +6753,11 @@ def sequence_scatter(input, index, updates, name=None):
the columns to update in each row of X.
Here is an example:
Given the following input:
.. code-block:: text
input.data = [[1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
[1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
[1.0, 1.0, 1.0, 1.0, 1.0, 1.0]]
......@@ -6743,7 +6770,9 @@ def sequence_scatter(input, index, updates, name=None):
updates.lod = [[ 0, 3, 8, 12]]
Then we have the output:
.. code-block:: text
out.data = [[1.3, 1.3, 1.4, 1.0, 1.0, 1.0],
[1.0, 1.0, 1.4, 1.3, 1.2, 1.1],
[1.0, 1.0, 1.3, 1.2, 1.4, 1.1]]
......@@ -6759,7 +6788,7 @@ def sequence_scatter(input, index, updates, name=None):
name (str|None): The output variable name. Default None.
Returns:
output (Variable): The output is a tensor with the same shape as input.
Variable: The output is a tensor with the same shape as input.
Examples:
......@@ -6933,7 +6962,7 @@ def mean_iou(input, label, num_classes):
.. math::
IOU = \\frac{true\_positiv}{(true\_positive + false\_positive + false\_negative)}.
IOU = \\frac{true\_positive}{(true\_positive + false\_positive + false\_negative)}.
The predictions are accumulated in a confusion matrix and mean-IOU
is then calculated from it.
......@@ -6946,9 +6975,13 @@ def mean_iou(input, label, num_classes):
num_classes (int): The possible number of labels.
Returns:
mean_iou (Variable): A Tensor representing the mean intersection-over-union with shape [1].
out_wrong(Variable): A Tensor with shape [num_classes]. The wrong numbers of each class.
out_correct(Variable): A Tensor with shape [num_classes]. The correct numbers of each class.
mean_iou (Variable),out_wrong(Variable),out_correct(Variable):
Three variables:
- mean_iou : A Tensor representing the mean intersection-over-union with shape [1].
- out_wrong: A Tensor with shape [num_classes]. The wrong numbers of each class.
- out_correct: A Tensor with shape [num_classes]. The correct numbers of each class.
Examples:
......@@ -7144,7 +7177,7 @@ def affine_grid(theta, out_shape, name=None):
Args:
theta (Variable): A batch of affine transform parameters with shape [N, 2, 3].
out_shape (Variable | list | tuple): The shape of target output with format [N, C, H, W].
out_shape can be a Variable or a list or tuple.
``out_shape`` can be a Variable or a list or tuple.
name(str|None): A name for this layer(optional). If set None, the layer
will be named automatically.
......@@ -7157,6 +7190,7 @@ def affine_grid(theta, out_shape, name=None):
Examples:
.. code-block:: python
theta = fluid.layers.data(name="x", shape=[2, 3], dtype="float32")
out_shape = fluid.layers.data(name="y", shape=[-1], dtype="float32")
data = fluid.layers.affine_grid(theta, out_shape)
......@@ -7192,9 +7226,10 @@ def affine_grid(theta, out_shape, name=None):
def rank_loss(label, left, right, name=None):
"""
**Rank loss layer for RankNet**
RankNet(http://icml.cc/2015/wp-content/uploads/2015/06/icml_ranking.pdf)
`RankNet <http://icml.cc/2015/wp-content/uploads/2015/06/icml_ranking.pdf>`_
is a pairwise ranking model with a training sample consisting of a pair
of documents, A and B. Label P indicates whether A is ranked higher than B
or not:
......@@ -7202,16 +7237,19 @@ def rank_loss(label, left, right, name=None):
P = {0, 1} or {0, 0.5, 1}, where 0.5 means that there is no information
about the rank of the input pair.
Rank loss layer takes three inputs: left (o_i), right (o_j) and
label (P_{i,j}). The inputs respectively represent RankNet's output scores
Rank loss layer takes three inputs: left ( :math:`o_i` ), right ( :math:`o_j` ) and
label ( :math:`P_{i,j}` ). The inputs respectively represent RankNet's output scores
for documents A and B and the value of label P. The following equation
computes rank loss C_{i,j} from the inputs:
$$
C_{i,j} = -\tilde{P_{ij}} * o_{i,j} + \log(1 + e^{o_{i,j}}) \\
o_{i,j} = o_i - o_j \\
\tilde{P_{i,j}} = \left \{0, 0.5, 1 \right \} \ or \ \left \{0, 1 \right \}
$$
.. math::
C_{i,j} &= -\\tilde{P_{ij}} * o_{i,j} + \log(1 + e^{o_{i,j}}) \\\\
o_{i,j} &= o_i - o_j \\\\
\\tilde{P_{i,j}} &= \\left \{0, 0.5, 1 \\right \} \ or \ \\left \{0, 1 \\right \}
Rank loss layer takes batch inputs with size batch_size (batch_size >= 1).
......@@ -7237,7 +7275,6 @@ def rank_loss(label, left, right, name=None):
right = fluid.layers.data(name="right", shape=[4, 1], dtype="float32")
out = fluid.layers.rank_loss(label, left, right)
"""
helper = LayerHelper('rank_loss', **locals())
......@@ -7269,7 +7306,7 @@ def margin_rank_loss(label, left, right, margin=0.1, name=None):
.. math::
rank\_loss &= max(0, -label * (left - right) + margin)
rank\_loss = max(0, -label * (left - right) + margin)
Args:
label (Variable): Indicates whether the left is ranked higher than the right or not.
......@@ -7278,12 +7315,17 @@ def margin_rank_loss(label, left, right, margin=0.1, name=None):
margin (float): Indicates the given margin.
name (str|None): A name for this layer (optional). If set None, the layer
will be named automatically.
Returns:
Variable: The ranking loss.
Raises:
ValueError: Any of label, left, and right is not a Variable.
Examples:
.. code-block:: python
label = fluid.layers.data(name="label", shape=[4, 1], dtype="float32")
left = fluid.layers.data(name="left", shape=[4, 1], dtype="float32")
right = fluid.layers.data(name="right", shape=[4, 1], dtype="float32")
......@@ -7587,7 +7629,8 @@ def prelu(x, mode, param_attr=None, name=None):
"""
Equation:
y = \max(0, x) + alpha * \min(0, x)
.. math::
y = \max(0, x) + \\alpha * \min(0, x)
Args:
x (Variable): The input tensor.
......@@ -7730,20 +7773,29 @@ def flatten(x, axis=1, name=None):
**Flatten layer**
Flattens the input tensor into a 2D matrix.
Examples:
For Example:
.. code-block:: text
Case 1:
Given
X.shape = (3, 100, 100, 4)
and
axis = 2
We get:
Out.shape = (3 * 100, 4 * 100)
Case 2:
Given
X.shape = (3, 100, 100, 4)
and
axis = 0
We get:
Out.shape = (1, 3 * 100 * 100 * 4)
......@@ -7759,9 +7811,9 @@ def flatten(x, axis=1, name=None):
will be named automatically.
Returns:
Variable: A 2D tensor with the contents of the input tensor, with input
dimensions up to axis flattened to the outer dimension of
the output and remaining input dimensions flattened into the
Variable: A 2D tensor with the contents of the input tensor, with input \
dimensions up to axis flattened to the outer dimension of \
the output and remaining input dimensions flattened into the \
inner dimension of the output.
Raises:
......@@ -7801,15 +7853,19 @@ def sequence_enumerate(input, win_size, pad_value=0, name=None):
The enumerated sequence has the same 1st dimension with variable `input`, and
the 2nd dimension is `win_size`, padded by `pad_value` if necessary in generation.
Examples:
.. code-block:: text
Case 1:
Input:
X.lod = [[0, 3, 5]]
X.data = [[1], [2], [3], [4], [5]]
X.dims = [5, 1]
Attrs:
win_size = 2
pad_value = 0
Output:
Out.lod = [[0, 3, 5]]
Out.data = [[1, 2], [2, 3], [3, 0], [4, 5], [5, 0]]
......@@ -8896,6 +8952,7 @@ def similarity_focus(input, axis, indexes, name=None):
SimilarityFocus Operator
Generate a similarity focus mask with the same shape of input using the following method:
1. Extract the 3-D tensor(here the first dimension is BatchSize) corresponding
to the axis according to the indexes. For example, if axis=1 and indexes=[a],
it will get the matrix T=X[:, a, :, :]. In this case, if the shape of input X
......@@ -8969,14 +9026,16 @@ def similarity_focus(input, axis, indexes, name=None):
indexes(list): Indicating the indexes of the selected dimension.
Returns:
Variable: A tensor variable with the same shape and same type
Variable: A tensor variable with the same shape and same type \
as the input.
Examples:
.. code-block:: python
data = fluid.layers.data(
name='data', shape=[2, 3, 2, 2], dtype='float32')
x = fluid.layers.layer_norm(input=data, axis=1, indexes=[0])
"""
helper = LayerHelper('similarity_focus', **locals())
# check attrs
......@@ -9055,6 +9114,7 @@ def hash(input, hash_size, num_hash=1, name=None):
Examples:
.. code-block:: python
word_dict = paddle.dataset.imdb.word_dict()
x = fluid.layers.data(shape[1], dtype='int32', lod_level=1)
out = fluid.layers.hash(input=x, num_hash=4, hash_size=1000)
......@@ -9075,13 +9135,15 @@ def hash(input, hash_size, num_hash=1, name=None):
def grid_sampler(x, grid, name=None):
"""
This operation samples input X by using bilinear interpolation based on
flow field grid, which is usually gennerated by affine_grid. The grid of
flow field grid, which is usually gennerated by :code:`affine_grid` . The grid of
shape [N, H, W, 2] is the concatenation of (grid_x, grid_y) coordinates
with shape [N, H, W] each, where grid_x is indexing the 4th dimension
(in width dimension) of input data x and grid_y is indexng the 3rd
dimention (in height dimension), finally results is the bilinear
interpolation value of 4 nearest corner points.
.. code-block:: text
Step 1:
Get (x, y) grid coordinates and scale to [0, H-1/W-1].
......@@ -9126,16 +9188,18 @@ def grid_sampler(x, grid, name=None):
name (str, default None): The name of this layer.
Returns:
out(Variable): Output of shape [N, C, H, W] data samples input X
Variable: Output of shape [N, C, H, W] data samples input X
using bilnear interpolation based on input grid.
Exmples:
Examples:
.. code-block:: python
x = fluid.layers.data(name='x', shape=[3, 10, 32, 32], dtype='float32')
theta = fluid.layers.data(name='theta', shape=[3, 2, 3], dtype='float32')
grid = fluid.layers.affine_grid(input=theta, size=[3, 10, 32, 32]})
out = fluid.layers.grid_sampler(x=x, grid=grid)
"""
helper = LayerHelper("grid_sampler", **locals())
......@@ -9203,19 +9267,19 @@ def add_position_encoding(input, alpha, beta, name=None):
"""
**Add Position Encoding Layer**
This layer accepts an input 3D-Tensor of shape [N x M x P], and return an
This layer accepts an input 3D-Tensor of shape [N x M x P], and returns an
output Tensor of shape [N x M x P] with positional encoding value.
Refer to `Attention Is All You Need<http://arxiv.org/pdf/1706.03762.pdf>`_ .
Refer to `Attention Is All You Need <http://arxiv.org/pdf/1706.03762.pdf>`_ .
.. math::
PE(pos, 2i) = \\sin{(pos / 10000^{2i / P})} \\\\
PE(pos, 2i + 1) = \\cos{(pos / 10000^{2i / P})} \\\\
Out(:, pos, i) = \\alpha * input(:, pos, i) + \\beta * PE(pos, i)
PE(pos, 2i) &= \\sin{(pos / 10000^{2i / P})} \\\\
PE(pos, 2i + 1) &= \\cos{(pos / 10000^{2i / P})} \\\\
Out(:, pos, i) &= \\alpha * input(:, pos, i) + \\beta * PE(pos, i)
Where:
* PE(pos, 2i): the increment for the number at even position
* PE(pos, 2i + 1): the increment for the number at odd position
- :math:`PE(pos, 2i)` : the increment for the number at even position
- :math:`PE(pos, 2i + 1)` : the increment for the number at odd position
Args:
input (Variable): 3-D input tensor with shape [N x M x P]
......@@ -9230,6 +9294,7 @@ def add_position_encoding(input, alpha, beta, name=None):
.. code-block:: python
position_tensor = fluid.layers.add_position_encoding(input=tensor)
"""
helper = LayerHelper('add_position_encoding', **locals())
dtype = helper.input_dtype()
......@@ -9262,13 +9327,13 @@ def bilinear_tensor_product(x,
For example:
.. math::
out{i} = x * W_{i} * {y^\mathrm{T}}, i=0,1,...,size-1
out_{i} = x * W_{i} * {y^\mathrm{T}}, i=0,1,...,size-1
In this formula:
- :math:`x`: the first input contains M elements, shape is [batch_size, M].
- :math:`y`: the second input contains N elements, shape is [batch_size, N].
- :math:`W_{i}`: the i-th learned weight, shape is [M, N]
- :math:`out{i}`: the i-th element of out, shape is [batch_size, size].
- :math:`out_{i}`: the i-th element of out, shape is [batch_size, size].
- :math:`y^\mathrm{T}`: the transpose of :math:`y_{2}`.
Args:
......
python/paddle/fluid/layers/tensor.py
浏览文件 @ 66ea7184
......@@ -393,9 +393,6 @@ def fill_constant_batch_size_like(input,
It also sets *stop_gradient* to True.
>>> data = fluid.layers.fill_constant_batch_size_like(
>>> input=like, shape=[1], value=0, dtype='int64')
Args:
input(${input_type}): ${input_comment}.
......@@ -411,6 +408,14 @@ def fill_constant_batch_size_like(input,
Returns:
${out_comment}.
Examples:
.. code-block:: python
data = fluid.layers.fill_constant_batch_size_like(
input=like, shape=[1], value=0, dtype='int64')
"""
helper = LayerHelper("fill_constant_batch_size_like", **locals())
out = helper.create_variable_for_type_inference(dtype=dtype)
......
python/paddle/fluid/metrics.py
浏览文件 @ 66ea7184
......@@ -362,7 +362,7 @@ class ChunkEvaluator(MetricBase):
compute the precision recall and F1-score using the accumulated counter
numbers.
For some basics of chunking, please refer to
'Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>'.
`Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>`_ .
ChunkEvalEvaluator computes the precision, recall, and F1-score of chunk detection,
and supports IOB, IOE, IOBES and IO (also known as plain) tagging schemes.
......@@ -391,6 +391,7 @@ class ChunkEvaluator(MetricBase):
def update(self, num_infer_chunks, num_label_chunks, num_correct_chunks):
"""
Update the states based on the layers.chunk_eval() ouputs.
Args:
num_infer_chunks(int|numpy.array): The number of chunks in Inference on the given minibatch.
num_label_chunks(int|numpy.array): The number of chunks in Label on the given mini-batch.
......@@ -450,9 +451,9 @@ class EditDistance(MetricBase):
distance, instance_error = distance_evaluator.eval()
In the above example:
'distance' is the average of the edit distance in a pass.
'instance_error' is the instance error rate in a pass.
- 'distance' is the average of the edit distance in a pass.
- 'instance_error' is the instance error rate in a pass.
"""
......@@ -567,12 +568,15 @@ class DetectionMAP(object):
Calculate the detection mean average precision (mAP).
The general steps are as follows:
1. calculate the true positive and false positive according to the input
of detection and labels.
2. calculate mAP value, support two versions: '11 point' and 'integral'.
Please get more information from the following articles:
https://sanchom.wordpress.com/tag/average-precision/
https://arxiv.org/abs/1512.02325
Args:
......@@ -615,8 +619,10 @@ class DetectionMAP(object):
In the above example:
'cur_map_v' is the mAP of current mini-batch.
'accum_map_v' is the accumulative mAP of one pass.
- 'cur_map_v' is the mAP of current mini-batch.
- 'accum_map_v' is the accumulative mAP of one pass.
"""
def __init__(self,
......
python/paddle/fluid/transpiler/distribute_transpiler.py
浏览文件 @ 66ea7184
......@@ -125,14 +125,23 @@ def slice_variable(var_list, slice_count, min_block_size):
class DistributeTranspilerConfig(object):
"""
Args:
slice_var_up (bool): Do Tensor slice for pservers, default is True.
split_method (PSDispatcher): RoundRobin or HashName can be used
try to choose the best method to balance loads for pservers.
min_block_size (int): Minimum splitted element number in block.
According:https://github.com/PaddlePaddle/Paddle/issues/8638#issuecomment-369912156
.. py:attribute:: slice_var_up (bool)
Do Tensor slice for pservers, default is True.
.. py:attribute:: split_method (PSDispatcher)
RoundRobin or HashName can be used.
Try to choose the best method to balance loads for pservers.
.. py:attribute:: min_block_size (int)
Minimum number of splitted elements in block.
According to : https://github.com/PaddlePaddle/Paddle/issues/8638#issuecomment-369912156
We can use bandwidth effiently when data size is larger than 2MB.If you
want to change it, please be sure you see the slice_variable function.
want to change it, please be sure you have read the slice_variable function.
"""
slice_var_up = True
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或