nn.py 181.2 KB
Newer Older
1
#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
D
dzhwinter 已提交
2
#
D
dzhwinter 已提交
3 4 5
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
D
dzhwinter 已提交
6
#
D
dzhwinter 已提交
7
#     http://www.apache.org/licenses/LICENSE-2.0
D
dzhwinter 已提交
8
#
D
dzhwinter 已提交
9 10 11 12 13
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
Y
Yu Yang 已提交
14
"""
15
All layers just related to the neural network. 
Y
Yu Yang 已提交
16 17 18 19 20
"""

from ..layer_helper import LayerHelper
from ..initializer import Normal, Constant
from ..framework import Variable
Y
yangyaming 已提交
21
from ..param_attr import ParamAttr
Y
yuyang18 已提交
22
from layer_function_generator import autodoc, templatedoc
Y
yangyaming 已提交
23
from tensor import concat
C
chengduoZH 已提交
24
import utils
Y
yuyang18 已提交
25
import random
F
fengjiayi 已提交
26
from .. import unique_name
Y
Yu Yang 已提交
27 28

__all__ = [
Y
ying 已提交
29 30 31
    'fc',
    'embedding',
    'dynamic_lstm',
Y
Yibing Liu 已提交
32
    'dynamic_lstmp',
G
guosheng 已提交
33
    'dynamic_gru',
Y
ying 已提交
34 35 36 37 38 39 40 41 42
    'gru_unit',
    'linear_chain_crf',
    'crf_decoding',
    'cos_sim',
    'cross_entropy',
    'square_error_cost',
    'chunk_eval',
    'sequence_conv',
    'conv2d',
Y
yuyang18 已提交
43
    'conv3d',
Y
ying 已提交
44
    'sequence_pool',
45 46
    'sequence_softmax',
    'softmax',
Y
ying 已提交
47
    'pool2d',
Y
yuyang18 已提交
48
    'pool3d',
Y
ying 已提交
49 50 51
    'batch_norm',
    'beam_search_decode',
    'conv2d_transpose',
Y
yuyang18 已提交
52
    'conv3d_transpose',
Y
ying 已提交
53 54 55 56 57 58
    'sequence_expand',
    'lstm_unit',
    'reduce_sum',
    'reduce_mean',
    'reduce_max',
    'reduce_min',
59
    'reduce_prod',
Y
ying 已提交
60 61 62 63
    'sequence_first_step',
    'sequence_last_step',
    'dropout',
    'split',
64 65
    'ctc_greedy_decoder',
    'edit_distance',
Y
ying 已提交
66 67
    'l2_normalize',
    'matmul',
Q
qingqing01 已提交
68
    'topk',
Y
ying 已提交
69 70
    'warpctc',
    'sequence_reshape',
71
    'transpose',
72
    'im2sequence',
73
    'nce',
Q
Qiao Longfei 已提交
74
    'beam_search',
75
    'row_conv',
76
    'multiplex',
G
guosheng 已提交
77
    'layer_norm',
78 79
    'softmax_with_cross_entropy',
    'smooth_l1',
80
    'one_hot',
Y
Yu Yang 已提交
81
    'autoincreased_step_counter',
C
caoying03 已提交
82
    'reshape',
Y
yangyaming 已提交
83
    'lod_reset',
D
dragonwarrior 已提交
84
    'lrn',
G
guosheng 已提交
85
    'pad',
86
    'label_smooth',
87
    'roi_pool',
W
whs 已提交
88
    'dice_loss',
F
fengjiayi 已提交
89 90
    'image_resize',
    'image_resize_short',
B
baiyf 已提交
91
    'resize_bilinear',
W
whs 已提交
92
    'gather',
93
    'random_crop',
Y
yuyang18 已提交
94 95 96
    'mean_iou',
    'relu',
    'log',
Y
Yu Yang 已提交
97 98 99 100 101 102 103 104
]


def fc(input,
       size,
       num_flatten_dims=1,
       param_attr=None,
       bias_attr=None,
105
       use_mkldnn=False,
Y
Yu Yang 已提交
106
       act=None,
J
Jacek Czaja 已提交
107
       is_test=False,
108
       name=None):
Y
Yu Yang 已提交
109
    """
110
    **Fully Connected Layer**
Y
Yu Yang 已提交
111

F
fengjiayi 已提交
112 113 114 115 116 117 118 119 120
    This function creates a fully connected layer in the network. It can take 
    multiple tensors as its inputs. It creates a variable called weights for 
    each input tensor, which represents a fully connected weight matrix from 
    each input unit to each output unit. The fully connected layer multiplies 
    each input tensor with its coresponding weight to produce an output Tensor. 
    If multiple input tensors are given, the results of multiple multiplications 
    will be sumed up. If bias_attr is not None, a bias variable will be created 
    and added to the output. Finally, if activation is not None, it will be applied 
    to the output as well.
C
caoying03 已提交
121

C
caoying03 已提交
122
    This process can be formulated as follows:
123 124 125

    .. math::

126
        Out = Act({\sum_{i=0}^{N-1}X_iW_i + b})
127 128 129

    In the above equation:

C
caoying03 已提交
130 131 132 133
    * :math:`N`: Number of the input.
    * :math:`X_i`: The input tensor.
    * :math:`W`: The weights created by this layer.
    * :math:`b`: The bias parameter created by this layer (if needed).
134
    * :math:`Act`: The activation function.
C
caoying03 已提交
135
    * :math:`Out`: The output tensor.
Y
Yu Yang 已提交
136 137

    Args:
R
ranqiu 已提交
138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154
        input (Variable|list of Variable): The input tensor(s) of this layer, and the dimension of
            the input tensor(s) is at least 2.
        size(int): The number of output units in this layer.
        num_flatten_dims (int, default 1): The fc layer can accept an input tensor with more than
            two dimensions. If this happens, the multidimensional tensor will first be flattened
            into a 2-dimensional matrix. The parameter `num_flatten_dims` determines how the input
            tensor is flattened: the first `num_flatten_dims` (inclusive, index starts from 1)
            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.
            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.
        bias_attr (ParamAttr|list of ParamAttr, default None): The parameter attribute for the bias
            of this layer. If it is set to None, no bias will be added to the output units.
        act (str, default None): Activation to be applied to the output of this layer.
J
Jacek Czaja 已提交
155
        is_test(bool): A flag indicating whether execution is in test phase.
M
mozga-intel 已提交
156 157
        use_mkldnn(bool): Use mkldnn kernel or not, it is valid only when the mkldnn
            library is installed. Default: False
R
ranqiu 已提交
158
        name (str, default None): The name of this layer.
Y
Yu Yang 已提交
159

160
    Returns:
F
fengjiayi 已提交
161
        Variable: The transformation result.
162 163

    Raises:
C
caoying03 已提交
164
        ValueError: If rank of the input tensor is less than 2.
165 166 167 168

    Examples:
        .. code-block:: python

F
fengjiayi 已提交
169
          data = fluid.layers.data(name="data", shape=[32, 32], dtype="float32")
170
          fc = fluid.layers.fc(input=data, size=1000, act="tanh")
Y
Yu Yang 已提交
171
    """
C
caoying03 已提交
172

C
caoying03 已提交
173
    helper = LayerHelper("fc", **locals())
Y
Yu Yang 已提交
174 175 176 177

    dtype = helper.input_dtype()

    mul_results = []
178 179
    for input_var, param_attr in helper.iter_inputs_and_params():
        input_shape = input_var.shape
Y
Yu Yang 已提交
180 181 182
        param_shape = [
            reduce(lambda a, b: a * b, input_shape[num_flatten_dims:], 1)
        ] + [size]
Y
ying 已提交
183

Y
Yu Yang 已提交
184
        w = helper.create_parameter(
185 186
            attr=param_attr, shape=param_shape, dtype=dtype, is_bias=False)
        tmp = helper.create_tmp_variable(dtype)
187
        helper.append_op(
188 189 190
            type="mul",
            inputs={"X": input_var,
                    "Y": w},
191
            outputs={"Out": tmp},
M
mozga-intel 已提交
192 193
            attrs={"x_num_col_dims": num_flatten_dims,
                   "y_num_col_dims": 1})
194 195 196 197
        mul_results.append(tmp)

    if len(mul_results) == 1:
        pre_bias = mul_results[0]
198
    else:
199 200 201 202 203 204 205
        pre_bias = helper.create_tmp_variable(dtype)
        helper.append_op(
            type="sum", inputs={"X": mul_results}, outputs={"Out": pre_bias})
    # add bias
    pre_activation = helper.append_bias_op(pre_bias, dim_start=num_flatten_dims)
    # add activation
    return helper.append_activation(pre_activation)
Y
Yu Yang 已提交
206 207


208 209 210
def embedding(input,
              size,
              is_sparse=False,
211
              is_distributed=False,
212 213 214
              padding_idx=None,
              param_attr=None,
              dtype='float32'):
Y
Yu Yang 已提交
215
    """
216 217
    **Embedding Layer**

218
    This layer is used to lookup embeddings of IDs, provided by :attr:`input`, in
219 220
    a lookup table. The result of this lookup is the embedding of each ID in the
    :attr:`input`.
221 222 223

    All the input variables are passed in as local variables to the LayerHelper
    constructor.
Y
Yu Yang 已提交
224 225

    Args:
226 227 228 229 230
        input(Variable): The tensor variable containing the IDs.
        size(tuple|list): The shape of the look up table parameter. It should
            have two elements which indicate the size of the dictionary of
            embeddings and the size of each embedding vector respectively.
        is_sparse(bool): The flag indicating whether to use sparse update.
231
        is_distributed(bool): Whether to run lookup table from remote parameter server.
232 233
        padding_idx(int|long|None): If :attr:`None`, it makes no effect to lookup.
            Otherwise the given :attr:`padding_idx` indicates padding the output
234
            with zeros whenever lookup encounters it in :attr:`input`. If
235
            :math:`padding_idx < 0`, the :attr:`padding_idx` to use in lookup is
236 237
            :math:`size[0] + dim`.
        param_attr(ParamAttr): Parameters for this layer
238
        dtype(np.dtype|core.VarDesc.VarType|str): The type of data : float32, float_16, int etc
Y
Yu Yang 已提交
239

240 241 242
    Returns:
        Variable: The tensor variable storing the embeddings of the \
                  supplied inputs.
Y
Yu Yang 已提交
243

244 245
    Examples:
        .. code-block:: python
Y
Yu Yang 已提交
246

C
chengduoZH 已提交
247
          dict_size = len(dataset.ids)
248
          data = fluid.layers.data(name='ids', shape=[32, 32], dtype='float32')
C
chengduoZH 已提交
249
          fc = fluid.layers.embedding(input=data, size=[dict_size, 16])
Y
Yu Yang 已提交
250 251 252 253 254 255
    """

    helper = LayerHelper('embedding', **locals())
    w = helper.create_parameter(
        attr=helper.param_attr, shape=size, dtype=dtype, is_bias=False)
    tmp = helper.create_tmp_variable(dtype)
256 257
    padding_idx = -1 if padding_idx is None else padding_idx if padding_idx >= 0 else (
        size[0] + padding_idx)
Y
Yu Yang 已提交
258 259 260 261 262
    helper.append_op(
        type='lookup_table',
        inputs={'Ids': input,
                'W': w},
        outputs={'Out': tmp},
263 264 265 266 267
        attrs={
            'is_sparse': is_sparse,
            'is_distributed': is_distributed,
            'padding_idx': padding_idx
        })
Y
Yu Yang 已提交
268 269 270
    return tmp


Y
yi.wu 已提交
271
@templatedoc(op_type="lstm")
Y
Yu Yang 已提交
272 273
def dynamic_lstm(input,
                 size,
Y
Yancey 已提交
274 275
                 h_0=None,
                 c_0=None,
Y
Yu Yang 已提交
276 277 278 279 280 281 282
                 param_attr=None,
                 bias_attr=None,
                 use_peepholes=True,
                 is_reverse=False,
                 gate_activation='sigmoid',
                 cell_activation='tanh',
                 candidate_activation='tanh',
283 284
                 dtype='float32',
                 name=None):
Y
Yibing Liu 已提交
285
    """
Y
yi.wu 已提交
286
    ${comment}
Y
Yibing Liu 已提交
287 288

    Args:
Y
yi.wu 已提交
289 290
        input (Variable): ${input_comment}
        size (int): 4 * hidden size.
Y
Yancey 已提交
291 292 293 294 295 296 297
        h_0(Variable): The initial hidden state is an optional input, default is zero.
                       This is a tensor with shape (N x D), where N is the
                       batch size and D is the hidden size.
        c_0(Variable): The initial cell state is an optional input, default is zero.
                       This is a tensor with shape (N x D), where N is the
                       batch size. `h_0` and `c_0` can be NULL but only at the same time.

298
        param_attr(ParamAttr|None): The parameter attribute for the learnable
299
                               hidden-hidden weights.
Y
Yibing Liu 已提交
300 301 302

                               - Weights = {:math:`W_{ch}, W_{ih}, \
                                                W_{fh}, W_{oh}`}
303 304
                               - The shape is (D x 4D), where D is the hidden
                                 size.
Y
yi.wu 已提交
305
        bias_attr (ParamAttr|None): The bias attribute for the learnable bias
306 307 308
                              weights, which contains two parts, input-hidden
                              bias weights and peephole connections weights if
                              setting `use_peepholes` to `True`.
Y
Yibing Liu 已提交
309

310
                              1. `use_peepholes = False`
Y
yi.wu 已提交
311 312
                                 - Biases = {:math:`b_c, b_i, b_f, b_o`}.
                                 - The shape is (1 x 4D).
313
                              2. `use_peepholes = True`
Y
yi.wu 已提交
314
                                 - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \
Y
Yibing Liu 已提交
315
                                                 W_{fc}, W_{oc}`}.
Y
yi.wu 已提交
316
                                 - The shape is (1 x 7D).
Y
yi.wu 已提交
317 318 319 320 321 322 323 324
        use_peepholes (bool): ${use_peepholes_comment}
        is_reverse (bool): ${is_reverse_comment}
        gate_activation (str): ${gate_activation_comment}
        cell_activation (str): ${cell_activation_comment}
        candidate_activation (str): ${candidate_activation_comment}
        dtype (str): Data type. Choices = ["float32", "float64"], default "float32".
        name (str|None): A name for this layer(optional). If set None, the layer
                         will be named automatically.
Y
Yibing Liu 已提交
325 326

    Returns:
Y
Yibing Liu 已提交
327 328
        tuple: The hidden state, and cell state of LSTM. The shape of both \
        is (T x D), and lod is the same with the `input`.
Y
Yibing Liu 已提交
329

Y
Yibing Liu 已提交
330
    Examples:
Y
Yibing Liu 已提交
331 332
        .. code-block:: python

Y
Yibing Liu 已提交
333 334
            hidden_dim = 512
            forward_proj = fluid.layers.fc(input=input_seq, size=hidden_dim * 4,
335
                                           act=None, bias_attr=None)
Y
Yibing Liu 已提交
336 337
            forward, _ = fluid.layers.dynamic_lstm(
                input=forward_proj, size=hidden_dim * 4, use_peepholes=False)
Y
Yibing Liu 已提交
338
    """
339

Y
Yu Yang 已提交
340 341 342 343 344 345 346 347 348 349 350 351 352 353
    helper = LayerHelper('lstm', **locals())
    size = size / 4
    weight = helper.create_parameter(
        attr=helper.param_attr, shape=[size, 4 * size], dtype=dtype)
    bias_size = [1, 7 * size]
    if not use_peepholes:
        bias_size[1] = 4 * size
    bias = helper.create_parameter(
        attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True)

    hidden = helper.create_tmp_variable(dtype)
    cell = helper.create_tmp_variable(dtype)
    batch_gate = helper.create_tmp_variable(dtype)
    batch_cell_pre_act = helper.create_tmp_variable(dtype)
Y
Yancey 已提交
354 355 356 357 358 359 360 361 362 363
    inputs = {'Input': input, 'Weight': weight, 'Bias': bias}
    batch_size = input.shape[0]
    if h_0:
        assert h_0.shape == (batch_size, size), \
            'The shape of h0 should be (batch_size, %d)' % size
        inputs['H0'] = h_0
    if c_0:
        assert c_0.shape == (batch_size, size), \
            'The shape of c0 should be (batch_size, %d)' % size
        inputs['C0'] = c_0
Y
Yu Yang 已提交
364 365 366

    helper.append_op(
        type='lstm',
Y
Yancey 已提交
367
        inputs=inputs,
Y
Yu Yang 已提交
368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
        outputs={
            'Hidden': hidden,
            'Cell': cell,
            'BatchGate': batch_gate,
            'BatchCellPreAct': batch_cell_pre_act
        },
        attrs={
            'use_peepholes': use_peepholes,
            'is_reverse': is_reverse,
            'gate_activation': gate_activation,
            'cell_activation': cell_activation,
            'candidate_activation': candidate_activation
        })
    return hidden, cell


Y
Yibing Liu 已提交
384 385 386 387 388 389 390 391 392 393 394
def dynamic_lstmp(input,
                  size,
                  proj_size,
                  param_attr=None,
                  bias_attr=None,
                  use_peepholes=True,
                  is_reverse=False,
                  gate_activation='sigmoid',
                  cell_activation='tanh',
                  candidate_activation='tanh',
                  proj_activation='tanh',
395 396
                  dtype='float32',
                  name=None):
Y
Yibing Liu 已提交
397 398 399
    """
    **Dynamic LSTMP Layer**

400 401 402 403 404 405
    LSTMP (LSTM with recurrent projection) layer has a separate projection
    layer after the LSTM layer, projecting the original hidden state to a
    lower-dimensional one, which is proposed to reduce the number of total
    parameters and furthermore computational complexity for the LSTM,
    espeacially for the case that the size of output units is relative
    large (https://research.google.com/pubs/archive/43905.pdf).
Y
Yibing Liu 已提交
406 407 408 409 410

    The formula is as follows:

    .. math::

411
        i_t & = \sigma(W_{ix}x_{t} + W_{ir}r_{t-1} + W_{ic}c_{t-1} + b_i)
Y
Yibing Liu 已提交
412

413
        f_t & = \sigma(W_{fx}x_{t} + W_{fr}r_{t-1} + W_{fc}c_{t-1} + b_f)
Y
Yibing Liu 已提交
414

415
        \\tilde{c_t} & = act_g(W_{cx}x_t + W_{cr}r_{t-1} + b_c)
Y
Yibing Liu 已提交
416

417
        o_t & = \sigma(W_{ox}x_{t} + W_{or}r_{t-1} + W_{oc}c_t + b_o)
Y
Yibing Liu 已提交
418

419
        c_t & = f_t \odot c_{t-1} + i_t \odot \\tilde{c_t}
Y
Yibing Liu 已提交
420

421
        h_t & = o_t \odot act_h(c_t)
Y
Yibing Liu 已提交
422

423
        r_t & = \overline{act_h}(W_{rh}h_t)
Y
Yibing Liu 已提交
424

Y
Yibing Liu 已提交
425 426 427 428 429 430
    In the above formula:

    * :math:`W`: Denotes weight matrices (e.g. :math:`W_{xi}` is \
          the matrix of weights from the input gate to the input).
    * :math:`W_{ic}`, :math:`W_{fc}`, :math:`W_{oc}`: Diagonal weight \
          matrices for peephole connections. In our implementation, \
431
          we use vectors to reprenset these diagonal weight matrices.
Y
Yibing Liu 已提交
432
    * :math:`b`: Denotes bias vectors (e.g. :math:`b_i` is the input gate \
433
          bias vector).
Y
Yibing Liu 已提交
434 435 436
    * :math:`\sigma`: The activation, such as logistic sigmoid function.
    * :math:`i, f, o` and :math:`c`: The input gate, forget gate, output \
          gate, and cell activation vectors, respectively, all of which have \
437
          the same size as the cell output activation vector :math:`h`.
Y
Yibing Liu 已提交
438
    * :math:`h`: The hidden state.
439
    * :math:`r`: The recurrent projection of the hidden state.
Y
Yibing Liu 已提交
440 441
    * :math:`\\tilde{c_t}`: The candidate hidden state, whose \
          computation is based on the current input and previous hidden state.
442
    * :math:`\odot`: The element-wise product of the vectors.
Y
Yibing Liu 已提交
443
    * :math:`act_g` and :math:`act_h`: The cell input and cell output \
444
          activation functions and `tanh` is usually used for them.
Y
Yibing Liu 已提交
445 446
    * :math:`\overline{act_h}`: The activation function for the projection \
          output, usually using `identity` or same as :math:`act_h`.
Y
Yibing Liu 已提交
447 448 449 450

    Set `use_peepholes` to `False` to disable peephole connection. The formula
    is omitted here, please refer to the paper
    http://www.bioinf.jku.at/publications/older/2604.pdf for details.
451

Y
Yibing Liu 已提交
452 453 454 455 456 457 458 459 460 461 462 463
    Note that these :math:`W_{xi}x_{t}, W_{xf}x_{t}, W_{xc}x_{t}, W_{xo}x_{t}`
    operations on the input :math:`x_{t}` are NOT included in this operator.
    Users can choose to use fully-connected layer before LSTMP layer.

    Args:
        input(Variable): The input of dynamic_lstmp layer, which supports
                         variable-time length input sequence. The underlying
                         tensor in this Variable is a matrix with shape
                         (T X 4D), where T is the total time steps in this
                         mini-batch, D is the hidden size.
        size(int): 4 * hidden size.
        proj_size(int): The size of projection output.
464
        param_attr(ParamAttr|None): The parameter attribute for the learnable
Y
Yibing Liu 已提交
465 466
                               hidden-hidden weight and projection weight.

467 468
                               - Hidden-hidden weight = {:math:`W_{ch}, W_{ih}, \
                                                W_{fh}, W_{oh}`}.
469 470
                               - The shape of hidden-hidden weight is (P x 4D),
                                 where P is the projection size and D the hidden
Y
Yibing Liu 已提交
471 472
                                 size.
                               - Projection weight = {:math:`W_{rh}`}.
473 474
                               - The shape of projection weight is (D x P).
        bias_attr(ParamAttr|None): The bias attribute for the learnable bias
Y
Yibing Liu 已提交
475 476 477 478 479 480
                              weights, which contains two parts, input-hidden
                              bias weights and peephole connections weights if
                              setting `use_peepholes` to `True`.

                              1. `use_peepholes = False`
                                - Biases = {:math:`b_c, b_i, b_f, b_o`}.
481
                                - The shape is (1 x 4D).
Y
Yibing Liu 已提交
482 483 484
                              2. `use_peepholes = True`
                                - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \
                                                 W_{fc}, W_{oc}`}.
485
                                - The shape is (1 x 7D).
Y
Yibing Liu 已提交
486 487 488 489 490 491 492 493 494
        use_peepholes(bool): Whether to enable diagonal/peephole connections,
                             default `True`.
        is_reverse(bool): Whether to compute reversed LSTM, default `False`.
        gate_activation(str): The activation for input gate, forget gate and
                              output gate. Choices = ["sigmoid", "tanh", "relu",
                              "identity"], default "sigmoid".
        cell_activation(str): The activation for cell output. Choices = ["sigmoid",
                              "tanh", "relu", "identity"], default "tanh".
        candidate_activation(str): The activation for candidate hidden state.
495
                              Choices = ["sigmoid", "tanh", "relu", "identity"],
Y
Yibing Liu 已提交
496 497
                              default "tanh".
        proj_activation(str): The activation for projection output.
498
                              Choices = ["sigmoid", "tanh", "relu", "identity"],
Y
Yibing Liu 已提交
499 500
                              default "tanh".
        dtype(str): Data type. Choices = ["float32", "float64"], default "float32".
501 502
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
Y
Yibing Liu 已提交
503 504

    Returns:
505 506 507 508
        tuple: A tuple of two output variable: the projection of hidden state, \
               and cell state of LSTMP. The shape of projection is (T x P), \
               for the cell state which is (T x D), and both LoD is the same \
               with the `input`.
Y
Yibing Liu 已提交
509 510

    Examples:
511

Y
Yibing Liu 已提交
512 513
        .. code-block:: python

514 515 516 517
            dict_dim, emb_dim = 128, 64
            data = fluid.layers.data(name='sequence', shape=[1],
                                     dtype='int32', lod_level=1)
            emb = fluid.layers.embedding(input=data, size=[dict_dim, emb_dim])
Y
Yibing Liu 已提交
518
            hidden_dim, proj_dim = 512, 256
519
            fc_out = fluid.layers.fc(input=emb, size=hidden_dim * 4,
Y
Yibing Liu 已提交
520
                                     act=None, bias_attr=None)
521 522 523
            proj_out, _ = fluid.layers.dynamic_lstmp(input=fc_out,
                                                     size=hidden_dim * 4,
                                                     proj_size=proj_dim,
Y
Yibing Liu 已提交
524 525 526 527
                                                     use_peepholes=False,
                                                     is_reverse=True,
                                                     cell_activation="tanh",
                                                     proj_activation="tanh")
Y
Yibing Liu 已提交
528
    """
529

Y
Yibing Liu 已提交
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
    helper = LayerHelper('lstmp', **locals())
    size = size / 4
    weight = helper.create_parameter(
        attr=helper.param_attr, shape=[proj_size, 4 * size], dtype=dtype)
    proj_weight = helper.create_parameter(
        attr=helper.param_attr, shape=[size, proj_size], dtype=dtype)
    bias_size = [1, 7 * size]
    if not use_peepholes:
        bias_size[1] = 4 * size
    bias = helper.create_parameter(
        attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True)

    projection = helper.create_tmp_variable(dtype)
    cell = helper.create_tmp_variable(dtype)
    ordered_proj0 = helper.create_tmp_variable(dtype)
    batch_hidden = helper.create_tmp_variable(dtype)
    batch_gate = helper.create_tmp_variable(dtype)
    batch_cell_pre_act = helper.create_tmp_variable(dtype)

    helper.append_op(
        type='lstmp',
        inputs={
            'Input': input,
            'Weight': weight,
            'ProjWeight': proj_weight,
            'Bias': bias
        },
        outputs={
            'Projection': projection,
            'Cell': cell,
            'OrderedP0': ordered_proj0,
            'BatchHidden': batch_hidden,
            'BatchGate': batch_gate,
            'BatchCellPreAct': batch_cell_pre_act
        },
        attrs={
            'use_peepholes': use_peepholes,
            'is_reverse': is_reverse,
            'gate_activation': gate_activation,
            'cell_activation': cell_activation,
            'candidate_activation': candidate_activation,
            'proj_activation': proj_activation
        })
    return projection, cell


G
guosheng 已提交
576 577 578 579 580 581 582 583 584
def dynamic_gru(input,
                size,
                param_attr=None,
                bias_attr=None,
                is_reverse=False,
                gate_activation='sigmoid',
                candidate_activation='tanh',
                h_0=None):
    """
585
    **Gated Recurrent Unit (GRU) Layer**
G
guosheng 已提交
586

587
    Refer to `Empirical Evaluation of Gated Recurrent Neural Networks on
588
    Sequence Modeling <https://arxiv.org/abs/1412.3555>`_ .
589

G
guosheng 已提交
590 591 592 593 594 595 596 597 598
    The formula is as follows:

    .. math::

        u_t & = act_g(W_{ux}x_{t} + W_{uh}h_{t-1} + b_u)

        r_t & = act_g(W_{rx}x_{t} + W_{rh}h_{t-1} + b_r)

        \\tilde{h_t} & = act_c(W_{cx}x_{t} + W_{ch}(r_t \odot h_{t-1}) + b_c)
599

G
guosheng 已提交
600
        h_t & = (1-u_t) \odot h_{t-1} + u_t \odot \\tilde{h_t}
601

G
guosheng 已提交
602
    The :math:`\odot` is the element-wise product of the vectors. :math:`act_g`
603 604
    is the update gate and reset gate activation function and :math:`sigmoid`
    is usually used for it. :math:`act_c` is the activation function for
G
guosheng 已提交
605 606 607 608
    candidate hidden state and :math:`tanh` is usually used for it.

    Note that these :math:`W_{ux}x_{t}, W_{rx}x_{t}, W_{cx}x_{t}` operations on
    the input :math:`x_{t}` are NOT included in this operator. Users can choose
609
    to use fully-connect layer before GRU layer.
G
guosheng 已提交
610 611

    Args:
612 613
        input(Variable): The input of dynamic_gru layer, which supports
            variable-time length input sequence. The underlying tensor in this
G
guosheng 已提交
614
            Variable is a matrix with shape :math:`(T \\times 3D)`, where
615
            :math:`T` is the total time steps in this mini-batch, :math:`D`
G
guosheng 已提交
616 617
            is the hidden size.
        size(int): The dimension of the gru cell.
618
        param_attr(ParamAttr|None): The parameter attribute for the learnable
G
guosheng 已提交
619 620
            hidden-hidden weight matrix. Note:

621
            - The shape of the weight matrix is :math:`(T \\times 3D)`, where
G
guosheng 已提交
622
              :math:`D` is the hidden size.
623
            - All elements in the weight matrix can be divided into two parts.
G
guosheng 已提交
624
              The first part are weights of the update gate and reset gate with
625
              shape :math:`(D \\times 2D)`, and the second part are weights for
G
guosheng 已提交
626
              candidate hidden state with shape :math:`(D \\times D)`.
627
        bias_attr(ParamAttr): The parameter attribute for learnable the
G
guosheng 已提交
628
            hidden-hidden bias.
629
        is_reverse(bool): Whether to compute reversed GRU, default
G
guosheng 已提交
630 631 632
            :attr:`False`.
        gate_activation(str): The activation for update gate and reset gate.
            Choices = ["sigmoid", "tanh", "relu", "identity"], default "sigmoid".
633
        candidate_activation(str): The activation for candidate hidden state.
G
guosheng 已提交
634
            Choices = ["sigmoid", "tanh", "relu", "identity"], default "tanh".
635 636 637 638
        h_0 (Variable): This is initial hidden state. If not set, default is
            zero. This is a tensor with shape (N x D), where N is the number of
            total time steps of input mini-batch feature and D is the hidden
            size.
G
guosheng 已提交
639 640

    Returns:
G
guosheng 已提交
641
        Variable: The hidden state of GRU. The shape is :math:`(T \\times D)`, \
642
            and sequence length is the same with the input.
643

G
guosheng 已提交
644
    Examples:
645

G
guosheng 已提交
646 647
        .. code-block:: python

648 649 650 651
            dict_dim, emb_dim = 128, 64
            data = fluid.layers.data(name='sequence', shape=[1],
                                     dtype='int32', lod_level=1)
            emb = fluid.layers.embedding(input=data, size=[dict_dim, emb_dim])
G
guosheng 已提交
652
            hidden_dim = 512
653
            x = fluid.layers.fc(input=emb, size=hidden_dim * 3)
G
guosheng 已提交
654 655 656 657 658 659 660 661 662 663
            hidden = fluid.layers.dynamic_gru(input=x, dim=hidden_dim)
    """

    helper = LayerHelper('gru', **locals())
    dtype = helper.input_dtype()

    weight = helper.create_parameter(
        attr=helper.param_attr, shape=[size, 3 * size], dtype=dtype)
    bias = helper.create_parameter(
        attr=helper.bias_attr, shape=[1, 3 * size], dtype=dtype, is_bias=True)
Y
Yancey 已提交
664
    batch_size = input.shape[0]
G
guosheng 已提交
665 666 667
    inputs = {'Input': input, 'Weight': weight, 'Bias': bias}
    if h_0 != None:
        assert h_0.shape == (
Y
Yancey 已提交
668 669 670
            batch_size, size
        ), 'The shape of h0 should be(batch_size, %d)' % size
        inputs['H0'] = h_0
G
guosheng 已提交
671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693

    hidden = helper.create_tmp_variable(dtype)
    batch_gate = helper.create_tmp_variable(dtype)
    batch_reset_hidden_prev = helper.create_tmp_variable(dtype)
    batch_hidden = helper.create_tmp_variable(dtype)

    helper.append_op(
        type='gru',
        inputs=inputs,
        outputs={
            'Hidden': hidden,
            'BatchGate': batch_gate,
            'BatchResetHiddenPrev': batch_reset_hidden_prev,
            'BatchHidden': batch_hidden
        },
        attrs={
            'is_reverse': is_reverse,
            'gate_activation': gate_activation,
            'activation': candidate_activation
        })
    return hidden


Y
Yu Yang 已提交
694 695 696
def gru_unit(input,
             hidden,
             size,
697 698
             param_attr=None,
             bias_attr=None,
Y
Yu Yang 已提交
699
             activation='tanh',
700
             gate_activation='sigmoid'):
Y
Yu Yang 已提交
701
    """
702
    GRU unit layer. The equation of a gru step is:
Y
Yu Yang 已提交
703

704 705
        .. math::
            u_t & = actGate(xu_{t} + W_u h_{t-1} + b_u)
Y
Yu Yang 已提交
706

707
            r_t & = actGate(xr_{t} + W_r h_{t-1} + b_r)
Y
Yu Yang 已提交
708

709
            m_t & = actNode(xm_t + W_c dot(r_t, h_{t-1}) + b_m)
710

711
            h_t & = dot((1-u_t), m_t) + dot(u_t, h_{t-1})
712 713

    The inputs of gru unit includes :math:`z_t`, :math:`h_{t-1}`. In terms
714 715 716
    of the equation above, the :math:`z_t` is split into 3 parts -
    :math:`xu_t`, :math:`xr_t` and :math:`xm_t`. This means that in order to
    implement a full GRU unit operator for an input, a fully
717 718
    connected layer has to be applied, such that :math:`z_t = W_{fc}x_t`.

719 720
    The terms :math:`u_t` and :math:`r_t` represent the update and reset gates
    of the GRU cell. Unlike LSTM, GRU has one lesser gate. However, there is
721 722 723
    an intermediate candidate hidden output, which is denoted by :math:`m_t`.
    This layer has three outputs :math:`h_t`, :math:`dot(r_t, h_{t-1})`
    and concatenation of :math:`u_t`, :math:`r_t` and :math:`m_t`.
724 725 726 727 728

    Args:
        input (Variable): The fc transformed input value of current step.
        hidden (Variable): The hidden value of lstm unit from previous step.
        size (integer): The input dimension value.
729 730
        param_attr (ParamAttr): The weight parameters for gru unit. Default: None
        bias_attr (ParamAttr): The bias parameters for gru unit. Default: None
731 732 733 734
        activation (string): The activation type for cell (actNode).
                             Default: 'tanh'
        gate_activation (string): The activation type for gates (actGate).
                                  Default: 'sigmoid'
Y
Yu Yang 已提交
735

736 737 738 739 740 741
    Returns:
        tuple: The hidden value, reset-hidden value and gate values.

    Examples:

        .. code-block:: python
Y
Yu Yang 已提交
742

743
             # assuming we have x_t_data and prev_hidden of size=10
744
             x_t = fluid.layers.fc(input=x_t_data, size=30)
745 746
             hidden_val, r_h_val, gate_val = fluid.layers.gru_unit(input=x_t,
                                                    hidden = prev_hidden)
Y
Yu Yang 已提交
747 748 749 750 751 752 753 754 755 756 757 758 759 760 761

    """
    activation_dict = dict(
        identity=0,
        sigmoid=1,
        tanh=2,
        relu=3, )
    activation = activation_dict[activation]
    gate_activation = activation_dict[gate_activation]

    helper = LayerHelper('gru_unit', **locals())
    dtype = helper.input_dtype()
    size = size / 3

    # create weight
762 763
    weight = helper.create_parameter(
        attr=helper.param_attr, shape=[size, 3 * size], dtype=dtype)
Y
Yu Yang 已提交
764

765 766 767 768
    gate = helper.create_tmp_variable(dtype)
    reset_hidden_pre = helper.create_tmp_variable(dtype)
    updated_hidden = helper.create_tmp_variable(dtype)
    inputs = {'Input': input, 'HiddenPrev': hidden, 'Weight': weight}
Y
Yu Yang 已提交
769
    # create bias
770
    if helper.bias_attr:
Y
Yu Yang 已提交
771 772 773
        bias_size = [1, 3 * size]
        bias = helper.create_parameter(
            attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True)
774
        inputs['Bias'] = bias
Y
Yu Yang 已提交
775 776 777

    helper.append_op(
        type='gru_unit',
778
        inputs=inputs,
Y
Yu Yang 已提交
779 780 781 782 783 784
        outputs={
            'Gate': gate,
            'ResetHiddenPrev': reset_hidden_pre,
            'Hidden': updated_hidden,
        },
        attrs={
785 786
            'activation': 2,  # tanh
            'gate_activation': 1,  # sigmoid
Y
Yu Yang 已提交
787 788 789 790 791
        })

    return updated_hidden, reset_hidden_pre, gate


Y
yuyang18 已提交
792
@templatedoc()
793
def linear_chain_crf(input, label, param_attr=None):
Y
yuyang18 已提交
794 795 796 797 798 799 800
    """
    Linear Chain CRF.

    ${comment}

    Args:
        input(${emission_type}): ${emission_comment}
D
dzhwinter 已提交
801
        input(${transition_type}): ${transition_comment}
Y
yuyang18 已提交
802 803 804 805
        label(${label_type}): ${label_comment}
        param_attr(ParamAttr): The attribute of the learnable parameter.

    Returns:
D
dzhwinter 已提交
806 807 808
        output(${emission_exps_type}): ${emission_exps_comment} \n
        output(${transition_exps_type}): ${transition_exps_comment} \n
        output(${log_likelihood_type}): ${log_likelihood_comment}
Y
yuyang18 已提交
809 810

    """
Y
Yu Yang 已提交
811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
    helper = LayerHelper('linear_chain_crf', **locals())
    size = input.shape[1]
    transition = helper.create_parameter(
        attr=helper.param_attr,
        shape=[size + 2, size],
        dtype=helper.input_dtype())
    alpha = helper.create_tmp_variable(dtype=helper.input_dtype())
    emission_exps = helper.create_tmp_variable(dtype=helper.input_dtype())
    transition_exps = helper.create_tmp_variable(dtype=helper.input_dtype())
    log_likelihood = helper.create_tmp_variable(dtype=helper.input_dtype())
    helper.append_op(
        type='linear_chain_crf',
        inputs={"Emission": [input],
                "Transition": transition,
                "Label": label},
        outputs={
            "Alpha": [alpha],
            "EmissionExps": [emission_exps],
            "TransitionExps": transition_exps,
            "LogLikelihood": log_likelihood
        })

    return log_likelihood


Y
yuyang18 已提交
836
@templatedoc()
837
def crf_decoding(input, param_attr, label=None):
Y
yuyang18 已提交
838 839 840 841 842
    """
    ${comment}

    Args:
        input(${emission_type}): ${emission_comment}
Y
yi.wu 已提交
843

Y
yuyang18 已提交
844
        param_attr(ParamAttr): The parameter attribute for training.
Y
yi.wu 已提交
845

Y
yuyang18 已提交
846 847 848
        label(${label_type}): ${label_comment}

    Returns:
Y
update  
yi.wu 已提交
849
        Variable: ${viterbi_path_comment}
F
fengjiayi 已提交
850

Y
yi.wu 已提交
851 852 853 854 855
    Examples:
        .. code-block:: python

           crf_decode = layers.crf_decoding(
                input=hidden, param_attr=ParamAttr(name="crfw"))
Y
yuyang18 已提交
856
    """
Y
Yu Yang 已提交
857 858 859 860 861 862 863 864 865 866 867 868 869
    helper = LayerHelper('crf_decoding', **locals())
    transition = helper.get_parameter(param_attr.name)
    viterbi_path = helper.create_tmp_variable(dtype=helper.input_dtype())
    helper.append_op(
        type='crf_decoding',
        inputs={"Emission": [input],
                "Transition": transition,
                "Label": label},
        outputs={"ViterbiPath": [viterbi_path]})

    return viterbi_path


Y
yi.wu 已提交
870
@templatedoc()
F
fengjiayi 已提交
871
def cos_sim(X, Y):
Y
Yu Yang 已提交
872
    """
Y
yi.wu 已提交
873 874 875
    ${comment}

    Args:
876 877
        X (Variable): ${x_comment}.
        Y (Variable): ${y_comment}.
F
fengjiayi 已提交
878

Y
yi.wu 已提交
879
    Returns:
880
        Variable: the output of cosine(X, Y).
Y
Yu Yang 已提交
881
    """
F
fengjiayi 已提交
882
    helper = LayerHelper('cos_sim', **locals())
Y
Yu Yang 已提交
883 884 885 886 887 888 889 890 891 892 893 894 895
    out = helper.create_tmp_variable(dtype=X.dtype)
    xnorm = helper.create_tmp_variable(dtype=X.dtype)
    ynorm = helper.create_tmp_variable(dtype=X.dtype)
    helper.append_op(
        type='cos_sim',
        inputs={'X': [X],
                'Y': [Y]},
        outputs={'Out': [out],
                 'XNorm': [xnorm],
                 'YNorm': [ynorm]})
    return out


896
def dropout(x, dropout_prob, is_test=False, seed=None, name=None):
897 898 899 900 901
    """
    Computes dropout.

    Drop or keep each element of `x` independently. Dropout is a regularization
    technique for reducing overfitting by preventing neuron co-adaption during
902
    training. The dropout operator randomly sets (according to the given dropout
903 904 905 906
    probability) the outputs of some units to zero, while others are remain
    unchanged.

    Args:
907 908
        x (Variable): The input tensor variable.
        dropout_prob (float): Probability of setting units to zero.
909 910 911 912 913 914 915
        is_test (bool): A flag indicating whether it is in test phrase or not.
        seed (int): A Python integer used to create random seeds. If this
                    parameter is set to None, a random seed is used.
                    NOTE: If an integer seed is given, always the same output
                    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.
916 917

    Returns:
918
        Variable: A tensor variable is the shape with `x`.
919 920

    Examples:
921

922 923
        .. code-block:: python

924 925
            x = fluid.layers.data(name="data", shape=[32, 32], dtype="float32")
            droped = fluid.layers.dropout(x, dropout_prob=0.5)
926 927
    """

F
fengjiayi 已提交
928
    helper = LayerHelper('dropout', **locals())
929 930 931 932 933 934 935
    out = helper.create_tmp_variable(dtype=x.dtype)
    mask = helper.create_tmp_variable(dtype=x.dtype, stop_gradient=True)
    helper.append_op(
        type='dropout',
        inputs={'X': [x]},
        outputs={'Out': [out],
                 'Mask': [mask]},
936 937 938 939 940 941
        attrs={
            'dropout_prob': dropout_prob,
            'is_test': is_test,
            'fix_seed': seed is not None,
            'seed': seed if seed is not None else 0
        })
942 943 944
    return out


F
fengjiayi 已提交
945
def cross_entropy(input, label, soft_label=False):
Y
Yu Yang 已提交
946
    """
Y
Yibing Liu 已提交
947 948
    **Cross Entropy Layer**

949 950 951
    This layer computes the cross entropy between `input` and `label`. It
    supports both standard cross-entropy and soft-label cross-entropy loss
    computation.
Y
Yibing Liu 已提交
952 953

    1) One-hot cross-entropy:
F
fengjiayi 已提交
954
        `soft_label = False`, `Label[i, 0]` indicates the class index for sample i:
Y
yangyaming 已提交
955

Y
Yibing Liu 已提交
956
        .. math::
Y
yangyaming 已提交
957

Y
Yibing Liu 已提交
958 959 960
            Y[i] = -\log(X[i, Label[i]])

    2) Soft-label cross-entropy:
F
fengjiayi 已提交
961 962
        `soft_label = True`, `Label[i, j]` indicates the soft label of class j
        for sample i:
Y
Yibing Liu 已提交
963 964 965 966 967

        .. math::

            Y[i] = \sum_j{-Label[i, j] * log(X[i, j])}

Y
Yibing Liu 已提交
968
       Please make sure that in this case the summation of each row of `label`
Y
Yibing Liu 已提交
969 970 971
       equals one.

    3) One-hot cross-entropy with vecterized `label`:
F
fengjiayi 已提交
972 973
         As a special case of 2), when each row of 'label' has only one
         non-zero element which is equal to 1, soft-label cross-entropy degenerates
Y
Yibing Liu 已提交
974
         to a one-hot cross-entropy with one-hot label representation.
Y
yangyaming 已提交
975

Y
Yibing Liu 已提交
976
    Args:
Y
yangyaming 已提交
977
        input (Variable|list):  a 2-D tensor with shape [N x D], where N is the
978 979 980 981
                                batch size and D is the number of classes. This
                                input is a probability computed by the previous
                                operator, which is almost always the result of
                                a softmax operator.
Y
yangyaming 已提交
982
        label (Variable|list): the ground truth which is a 2-D tensor. When
983 984 985 986
                               `soft_label` is set to `False`, `label` is a
                               tensor<int64> with shape [N x 1]. When
                               `soft_label` is set to `True`, `label` is a
                               tensor<float/double> with shape [N x D].
F
fengjiayi 已提交
987
        soft_label (bool): a flag indicating whether to
988 989
                                           interpretate the given labels as soft
                                           labels, default `False`.
Y
Yibing Liu 已提交
990 991 992 993 994

    Returns:
         A 2-D tensor with shape [N x 1], the cross entropy loss.

    Raises:
995 996 997 998 999
        `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.
Y
Yibing Liu 已提交
1000 1001 1002 1003 1004 1005

    Examples:
        .. code-block:: python

          predict = fluid.layers.fc(input=net, size=classdim, act='softmax')
          cost = fluid.layers.cross_entropy(input=predict, label=label)
Y
Yu Yang 已提交
1006
    """
F
fengjiayi 已提交
1007
    helper = LayerHelper('cross_entropy', **locals())
Y
Yu Yang 已提交
1008 1009 1010 1011 1012 1013
    out = helper.create_tmp_variable(dtype=input.dtype)
    helper.append_op(
        type='cross_entropy',
        inputs={'X': [input],
                'Label': [label]},
        outputs={'Y': [out]},
F
fengjiayi 已提交
1014
        attrs={"soft_label": soft_label})
Y
Yu Yang 已提交
1015 1016 1017
    return out


F
fengjiayi 已提交
1018
def square_error_cost(input, label):
Y
Yu Yang 已提交
1019
    """
1020 1021
    **Square error cost layer**

1022 1023
    This layer accepts input predictions and target label and returns the
    squared error cost.
Y
ying 已提交
1024

1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
    For predictions, :math:`X`, and target labels, :math:`Y`, the equation is:

    .. math::

        Out = (X - Y)^2

    In the above equation:

        * :math:`X`: Input predictions, a tensor.
        * :math:`Y`: Input labels, a tensor.
        * :math:`Out`: Output value, same shape with :math:`X`.

    Args:
1038 1039
        input (Variable): Input tensor, has predictions.
        label (Variable): Label tensor, has target labels.
1040 1041

    Returns:
G
guosheng 已提交
1042
        Variable: The tensor variable storing the element-wise squared error \
1043
                  difference of input and label.
1044 1045 1046 1047 1048 1049 1050 1051

    Examples:
        .. code-block:: python

          y = layers.data(name='y', shape=[1], dtype='float32')
          y_predict = layers.data(name='y_predict', shape=[1], dtype='float32')
          cost = layers.square_error_cost(input=y_predict, label=y)

Y
Yu Yang 已提交
1052
    """
F
fengjiayi 已提交
1053
    helper = LayerHelper('square_error_cost', **locals())
Y
Yu Yang 已提交
1054 1055 1056 1057 1058 1059 1060 1061 1062
    minus_out = helper.create_tmp_variable(dtype=input.dtype)
    helper.append_op(
        type='elementwise_sub',
        inputs={'X': [input],
                'Y': [label]},
        outputs={'Out': [minus_out]})

    square_out = helper.create_tmp_variable(dtype=input.dtype)
    helper.append_op(
F
fengjiayi 已提交
1063 1064
        type='square', inputs={'X': [minus_out]},
        outputs={'Out': [square_out]})
Y
Yu Yang 已提交
1065 1066 1067
    return square_out


Y
yi.wu 已提交
1068
@templatedoc()
Y
Yu Yang 已提交
1069 1070 1071 1072
def chunk_eval(input,
               label,
               chunk_scheme,
               num_chunk_types,
F
fengjiayi 已提交
1073
               excluded_chunk_types=None):
Y
Yu Yang 已提交
1074
    """
Y
yi.wu 已提交
1075
    **Chunk Evaluator**
Y
yi.wu 已提交
1076

Y
yangyaming 已提交
1077
    This function computes and outputs the precision, recall and
1078
    F1-score of chunk detection.
Y
yi.wu 已提交
1079

Y
yi.wu 已提交
1080 1081 1082 1083 1084 1085 1086 1087
    For some basics of chunking, please refer to
    '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.
    Here is a NER example of labeling for these tagging schemes:

    .. code-block:: python
F
fengjiayi 已提交
1088

Y
yi.wu 已提交
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113
       ====== ====== ======  =====  ==  ============   =====  ===== =====  ==  =========
              Li     Ming    works  at  Agricultural   Bank   of    China  in  Beijing.
       ====== ====== ======  =====  ==  ============   =====  ===== =====  ==  =========
       IO     I-PER  I-PER   O      O   I-ORG          I-ORG  I-ORG I-ORG  O   I-LOC
       IOB    B-PER  I-PER   O      O   B-ORG          I-ORG  I-ORG I-ORG  O   B-LOC
       IOE    I-PER  E-PER   O      O   I-ORG          I-ORG  I-ORG E-ORG  O   E-LOC
       IOBES  B-PER  E-PER   O      O   I-ORG          I-ORG  I-ORG E-ORG  O   S-LOC
       ====== ====== ======  =====  ==  ============   =====  ===== =====  ==  =========

    There are three chunk types(named entity types) including PER(person), ORG(organization)
    and LOC(LOCATION), and we can see that the labels have the form <tag type>-<chunk type>.

    Since the calculations actually use label ids rather than labels, extra attention
    should be paid when mapping labels to ids to make CheckEvalOp work. The key point
    is that the listed equations are satisfied by ids.

    .. code-block:: python

       tag_type = label % num_tag_type
       chunk_type = label / num_tag_type

    where `num_tag_type` is the num of tag types in the tagging scheme, `num_chunk_type`
    is the num of chunk types, and `tag_type` get its value from the following table.

    .. code-block:: python
F
fengjiayi 已提交
1114

Y
yi.wu 已提交
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
       Scheme Begin Inside End   Single
        plain   0     -      -     -
        IOB     0     1      -     -
        IOE     -     0      1     -
        IOBES   0     1      2     3

    Still use NER as example, assuming the tagging scheme is IOB while chunk types are ORG,
    PER and LOC. To satisfy the above equations, the label map can be like this:

    .. code-block:: python

       B-ORG  0
       I-ORG  1
       B-PER  2
       I-PER  3
       B-LOC  4
       I-LOC  5
       O      6

    It's not hard to verify the equations noting that the num of chunk types
    is 3 and the num of tag types in IOB scheme is 2. For example, the label
    id of I-LOC is 5, the tag type id of I-LOC is 1, and the chunk type id of
    I-LOC is 2, which consistent with the results from the equations.

Y
yi.wu 已提交
1139
    Args:
1140 1141 1142 1143 1144
        input (Variable): prediction output of the network.
        label (Variable): label of the test data set.
        chunk_scheme (str): ${chunk_scheme_comment}
        num_chunk_types (int): ${num_chunk_types_comment}
        excluded_chunk_types (list): ${excluded_chunk_types_comment}
F
fengjiayi 已提交
1145

Y
yi.wu 已提交
1146
    Returns:
Y
update  
yi.wu 已提交
1147 1148 1149
        tuple: tuple containing: precision, recall, f1_score,
        num_infer_chunks, num_label_chunks,
        num_correct_chunks
F
fengjiayi 已提交
1150

Y
yi.wu 已提交
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
    Examples:
        .. code-block:: python

            crf = fluid.layers.linear_chain_crf(
                input=hidden, label=label, param_attr=ParamAttr(name="crfw"))
            crf_decode = fluid.layers.crf_decoding(
                input=hidden, param_attr=ParamAttr(name="crfw"))
            fluid.layers.chunk_eval(
                input=crf_decode,
                label=label,
                chunk_scheme="IOB",
                num_chunk_types=(label_dict_len - 1) / 2)
Y
Yu Yang 已提交
1163
    """
F
fengjiayi 已提交
1164
    helper = LayerHelper("chunk_eval", **locals())
Y
Yu Yang 已提交
1165 1166 1167 1168 1169

    # prepare output
    precision = helper.create_tmp_variable(dtype="float32")
    recall = helper.create_tmp_variable(dtype="float32")
    f1_score = helper.create_tmp_variable(dtype="float32")
1170 1171 1172
    num_infer_chunks = helper.create_tmp_variable(dtype="int64")
    num_label_chunks = helper.create_tmp_variable(dtype="int64")
    num_correct_chunks = helper.create_tmp_variable(dtype="int64")
Y
Yu Yang 已提交
1173 1174 1175 1176 1177 1178 1179 1180

    helper.append_op(
        type="chunk_eval",
        inputs={"Inference": [input],
                "Label": [label]},
        outputs={
            "Precision": [precision],
            "Recall": [recall],
1181 1182 1183 1184
            "F1-Score": [f1_score],
            "NumInferChunks": [num_infer_chunks],
            "NumLabelChunks": [num_label_chunks],
            "NumCorrectChunks": [num_correct_chunks]
Y
Yu Yang 已提交
1185 1186 1187
        },
        attrs={
            "num_chunk_types": num_chunk_types,
G
guosheng 已提交
1188 1189
            "chunk_scheme": chunk_scheme,
            "excluded_chunk_types": excluded_chunk_types or []
Y
Yu Yang 已提交
1190
        })
1191 1192
    return (precision, recall, f1_score, num_infer_chunks, num_label_chunks,
            num_correct_chunks)
Y
Yu Yang 已提交
1193 1194


1195
@templatedoc()
Y
Yu Yang 已提交
1196 1197 1198 1199 1200 1201 1202
def sequence_conv(input,
                  num_filters,
                  filter_size=3,
                  filter_stride=1,
                  padding=None,
                  bias_attr=None,
                  param_attr=None,
1203
                  act=None):
Y
Yu Yang 已提交
1204 1205 1206 1207
    """
    This function creates the op for sequence_conv, using the inputs and
    other convolutional configurations for the filters and stride as given
    in the input parameters to the function.
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217

    Args:
        input (Variable): ${x_comment}
        num_filters (int): number of filters.
        filter_size (int): the filter size (H and W).
        filter_stride (int): stride of the filter.
        padding (bool): if True, add paddings.
        bias_attr (ParamAttr|None): attributes for bias
        param_attr (ParamAttr|None): attributes for parameter
        act (str): the activation type
F
fengjiayi 已提交
1218

1219 1220
    Returns:
        Variable: output of sequence_conv
Y
Yu Yang 已提交
1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
    """

    helper = LayerHelper('sequence_conv', **locals())
    dtype = helper.input_dtype()
    filter_shape = [filter_size * input.shape[1], num_filters]
    filter_param = helper.create_parameter(
        attr=helper.param_attr, shape=filter_shape, dtype=dtype)
    pre_bias = helper.create_tmp_variable(dtype)

    helper.append_op(
        type='sequence_conv',
        inputs={
            'X': [input],
            'Filter': [filter_param],
        },
        outputs={"Out": pre_bias},
        attrs={
            'contextStride': filter_stride,
            'contextStart': -int(filter_size / 2),
            'contextLength': filter_size
        })
    pre_act = helper.append_bias_op(pre_bias)
    return helper.append_activation(pre_act)


1246
def sequence_softmax(input, param_attr=None, bias_attr=None, use_cudnn=True):
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
    """
    This function computes the softmax activation among all time-steps for each
    sequence. The dimension of each time-step should be 1. Thus, the shape of
    input Tensor can be either :math:`[N, 1]` or :math:`[N]`, where :math:`N` 
    is the sum of the length of all sequences.

    For i-th sequence in a mini-batch:

    .. math::

        Out(X[lod[i]:lod[i+1]], :) = \\frac{\exp(X[lod[i]:lod[i+1], :])}{\sum(\exp(X[lod[i]:lod[i+1], :]))}

    For example, for a mini-batch of 3 sequences with variable-length,
    each containing 2, 3, 2 time-steps, the lod of which is [0, 2, 5, 7],
    then softmax will be computed among :math:`X[0:2, :]`, :math:`X[2:5, :]`,
    :math:`X[5:7, :]`, and :math:`N` turns out to be 7.

    Args:
        input (Variable): The input variable which is a LoDTensor.
        bias_attr (ParamAttr|None): attributes for bias
        param_attr (ParamAttr|None): attributes for parameter
        use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn \
        library is installed. Default: True
F
fengjiayi 已提交
1270

1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
    Returns:
        Variable: output of sequence_softmax

    Examples:

        .. code-block:: python

             x = fluid.layers.data(name='x', shape=[7, 1],
                              dtype='float32', lod_level=1)
             x_sequence_softmax = fluid.layers.sequence_softmax(input=x)
    """
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
    helper = LayerHelper('sequence_softmax', **locals())
    dtype = helper.input_dtype()
    softmax_out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type="sequence_softmax",
        inputs={"X": input},
        outputs={"Out": softmax_out},
        attrs={"use_cudnn": use_cudnn})
    return softmax_out


1293
def softmax(input, param_attr=None, bias_attr=None, use_cudnn=True, name=None):
Q
qiaolongfei 已提交
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
    """
    The input of the softmax layer is a 2-D tensor with shape N x K (N is the
    batch_size, K is the dimension of input feature). The output tensor has the
    same shape as the input tensor.

    For each row of the input tensor, the softmax operator squashes the
    K-dimensional vector of arbitrary real values to a K-dimensional vector of real
    values in the range [0, 1] that add up to 1.

    It computes the exponential of the given dimension and the sum of exponential
    values of all the other dimensions in the K-dimensional vector input.
    Then the ratio of the exponential of the given dimension and the sum of
    exponential values of all the other dimensions is the output of the softmax
    operator.

    For each row :math:`i` and each column :math:`j` in Input(X), we have:

    .. math::

        Out[i, j] = \\frac{\exp(X[i, j])}{\sum_j(exp(X[i, j])}

    Args:
        input (Variable): The input variable.
        bias_attr (ParamAttr): attributes for bias
        param_attr (ParamAttr): attributes for parameter
        use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn \
        library is installed.

    Returns:
        Variable: output of softmax

    Examples:

        .. code-block:: python

             fc = fluid.layers.fc(input=x, size=10)
             softmax = fluid.layers.softmax(input=fc)

    """
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
    helper = LayerHelper('softmax', **locals())
    dtype = helper.input_dtype()
    softmax_out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type="softmax",
        inputs={"X": input},
        outputs={"Out": softmax_out},
        attrs={"use_cudnn": use_cudnn})
    return softmax_out


Y
Yu Yang 已提交
1344 1345 1346
def conv2d(input,
           num_filters,
           filter_size,
C
chengduoZH 已提交
1347 1348
           stride=1,
           padding=0,
1349
           dilation=1,
Y
Yu Yang 已提交
1350 1351 1352
           groups=None,
           param_attr=None,
           bias_attr=None,
C
chengduoZH 已提交
1353
           use_cudnn=True,
1354
           use_mkldnn=False,
1355 1356
           act=None,
           name=None):
Y
Yu Yang 已提交
1357
    """
C
chengduoZH 已提交
1358
    The convolution2D layer calculates the output based on the input, filter
T
tensor-tang 已提交
1359 1360
    and strides, paddings, dilations, groups parameters. Input and
    Output are in NCHW format, where N is batch size, C is the number of
1361
    channels, H is the height of the feature, and W is the width of the feature.
T
tensor-tang 已提交
1362 1363 1364 1365 1366 1367 1368
    Filter is in MCHW format, where M is the number of output image channels,
    C is the number of input image channels, H is the height of the filter,
    and W is the width of the filter. If the groups is greater than 1,
    C will equal the number of input image channels divided by the groups.
    Please refer to UFLDL's `convolution
    <http://ufldl.stanford.edu/tutorial/supervised/FeatureExtractionUsingConvolution/>`_
    for more detials.
1369 1370 1371
    If bias attribution and activation type are provided, bias is added to the
    output of the convolution, and the corresponding activation function is
    applied to the final result.
C
chengduoZH 已提交
1372

1373
    For each input :math:`X`, the equation is:
C
refine  
chengduoZH 已提交
1374

C
chengduoZH 已提交
1375 1376
    .. math::

C
refine  
chengduoZH 已提交
1377
        Out = \sigma (W \\ast X + b)
C
chengduoZH 已提交
1378

T
tensor-tang 已提交
1379
    Where:
C
chengduoZH 已提交
1380

1381 1382 1383 1384 1385
    * :math:`X`: Input value, a tensor with NCHW format.
    * :math:`W`: Filter value, a tensor with MCHW format.
    * :math:`\\ast`: Convolution operation.
    * :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
    * :math:`\\sigma`: Activation function.
T
tensor-tang 已提交
1386
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
C
chengduoZH 已提交
1387 1388 1389

    Example:

1390 1391
        - Input:

W
weixing02 已提交
1392
          Input shape: :math:`(N, C_{in}, H_{in}, W_{in})`
C
refine  
chengduoZH 已提交
1393

W
weixing02 已提交
1394
          Filter shape: :math:`(C_{out}, C_{in}, H_f, W_f)`
C
refine  
chengduoZH 已提交
1395

1396
        - Output:
T
tensor-tang 已提交
1397

W
weixing02 已提交
1398
          Output shape: :math:`(N, C_{out}, H_{out}, W_{out})`
C
refine  
chengduoZH 已提交
1399

C
chengduoZH 已提交
1400
        Where
1401 1402

        .. math::
C
chengduoZH 已提交
1403

W
weixing02 已提交
1404 1405
            H_{out}&= \\frac{(H_{in} + 2 * paddings[0] - (dilations[0] * (H_f - 1) + 1))}{strides[0]} + 1 \\\\
            W_{out}&= \\frac{(W_{in} + 2 * paddings[1] - (dilations[1] * (W_f - 1) + 1))}{strides[1]} + 1
C
chengduoZH 已提交
1406 1407

    Args:
1408
        input (Variable): The input image with [N, C, H, W] format.
T
tensor-tang 已提交
1409
        num_filters(int): The number of filter. It is as same as the output
1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
            image channel.
        filter_size (int|tuple|None): The filter size. If filter_size is a tuple,
            it must contain two integers, (filter_size_H, filter_size_W).
            Otherwise, the filter will be a square.
        stride (int|tuple): The stride size. If stride is a tuple, it must
            contain two integers, (stride_H, stride_W). Otherwise, the
            stride_H = stride_W = stride. Default: stride = 1.
        padding (int|tuple): The padding size. If padding is a tuple, it must
            contain two integers, (padding_H, padding_W). Otherwise, the
            padding_H = padding_W = padding. Default: padding = 0.
        dilation (int|tuple): The dilation size. If dilation is a tuple, it must
            contain two integers, (dilation_H, dilation_W). Otherwise, the
            dilation_H = dilation_W = dilation. Default: dilation = 1.
        groups (int): The groups number of the Conv2d Layer. According to grouped
            convolution in Alex Krizhevsky's Deep CNN paper: when group=2,
            the first half of the filters is only connected to the first half
            of the input channels, while the second half of the filters is only
            connected to the second half of the input channels. Default: groups=1
        param_attr (ParamAttr): The parameters to the Conv2d Layer. Default: None
        bias_attr (ParamAttr): Bias parameter for the Conv2d layer. Default: None
        use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn
            library is installed. Default: True
T
tensor-tang 已提交
1432 1433
        use_mkldnn (bool): Use mkldnn kernels or not, it is valid only when compiled
            with mkldnn library. Default: False
1434 1435 1436
        act (str): Activation type. Default: None
        name (str|None): A name for this layer(optional). If set None, the layer
            will be named automatically.
C
chengduoZH 已提交
1437 1438

    Returns:
G
guosheng 已提交
1439
        Variable: The tensor variable storing the convolution and \
C
chengduoZH 已提交
1440 1441
                  non-linearity activation result.

C
refine  
chengduoZH 已提交
1442
    Raises:
1443 1444
        ValueError: If the shapes of input, filter_size, stride, padding and
                    groups mismatch.
C
refine  
chengduoZH 已提交
1445

C
chengduoZH 已提交
1446 1447 1448
    Examples:
        .. code-block:: python

1449 1450
          data = fluid.layers.data(name='data', shape=[3, 32, 32], dtype='float32')
          conv2d = fluid.layers.conv2d(input=data, num_filters=2, filter_size=3, act="relu")
Y
Yu Yang 已提交
1451 1452 1453
    """

    num_channels = input.shape[1]
1454 1455

    l_type = 'conv2d'
X
xzl 已提交
1456 1457
    if (num_channels == groups and num_filters % num_channels == 0 and
            not use_cudnn):
1458
        l_type = 'depthwise_conv2d'
1459 1460 1461 1462

    helper = LayerHelper(l_type, **locals())
    dtype = helper.input_dtype()

Y
Yu Yang 已提交
1463 1464 1465 1466 1467 1468 1469
    if groups is None:
        num_filter_channels = num_channels
    else:
        if num_channels % groups != 0:
            raise ValueError("num_channels must be divisible by groups.")
        num_filter_channels = num_channels / groups

C
chengduoZH 已提交
1470 1471 1472
    filter_size = utils.convert_to_list(filter_size, 2, 'filter_size')
    stride = utils.convert_to_list(stride, 2, 'stride')
    padding = utils.convert_to_list(padding, 2, 'padding')
1473
    dilation = utils.convert_to_list(dilation, 2, 'dilation')
C
chengduoZH 已提交
1474

C
chengduoZH 已提交
1475 1476
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
Y
Yu Yang 已提交
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493

    input_shape = input.shape
    filter_shape = [num_filters, num_filter_channels] + filter_size

    def _get_default_param_initializer():
        std = (2.0 / (filter_size[0]**2 * num_channels))**0.5
        return Normal(0.0, std, 0)

    filter_param = helper.create_parameter(
        attr=helper.param_attr,
        shape=filter_shape,
        dtype=dtype,
        default_initializer=_get_default_param_initializer())

    pre_bias = helper.create_tmp_variable(dtype)

    helper.append_op(
1494
        type=l_type,
Y
Yu Yang 已提交
1495 1496 1497 1498 1499
        inputs={
            'Input': input,
            'Filter': filter_param,
        },
        outputs={"Output": pre_bias},
C
chengduoZH 已提交
1500 1501 1502
        attrs={
            'strides': stride,
            'paddings': padding,
1503
            'dilations': dilation,
C
chengduoZH 已提交
1504
            'groups': groups,
1505 1506
            'use_cudnn': use_cudnn,
            'use_mkldnn': use_mkldnn
C
chengduoZH 已提交
1507
        })
Y
Yu Yang 已提交
1508 1509 1510 1511 1512 1513

    pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)

    return helper.append_activation(pre_act)


C
chengduoZH 已提交
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
def conv3d(input,
           num_filters,
           filter_size,
           stride=1,
           padding=0,
           dilation=1,
           groups=None,
           param_attr=None,
           bias_attr=None,
           use_cudnn=True,
           use_mkldnn=False,
           act=None,
           name=None):
    """
    **Convlution3D Layer**

    The convolution3D layer calculates the output based on the input, filter
    and strides, paddings, dilations, groups parameters. Input(Input) and
1532 1533 1534 1535 1536 1537
    Output(Output) are in NCDHW format. Where N is batch size C is the number of
    channels, D is the depth of the feature, H is the height of the feature,
    and W is the width of the feature. Convlution3D is similar with Convlution2D
    but adds one dimension(depth). If bias attribution and activation type are
    provided, bias is added to the output of the convolution, and the
    corresponding activation function is applied to the final result.
C
chengduoZH 已提交
1538 1539 1540 1541 1542 1543 1544 1545 1546

    For each input :math:`X`, the equation is:

    .. math::

        Out = \sigma (W \\ast X + b)

    In the above equation:

1547 1548
    * :math:`X`: Input value, a tensor with NCDHW format.
    * :math:`W`: Filter value, a tensor with MCDHW format.
C
chengduoZH 已提交
1549 1550 1551
    * :math:`\\ast`: Convolution operation.
    * :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
    * :math:`\\sigma`: Activation function.
1552
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
C
chengduoZH 已提交
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577

    Example:

        - Input:

          Input shape: :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`

          Filter shape: :math:`(C_{out}, C_{in}, D_f, H_f, W_f)`

        - Output:
          Output shape: :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`

        Where

        .. math::

            D_{out}&= \\frac{(D_{in} + 2 * paddings[0] - (dilations[0] * (D_f - 1) + 1))}{strides[0]} + 1 \\\\
            H_{out}&= \\frac{(H_{in} + 2 * paddings[1] - (dilations[1] * (H_f - 1) + 1))}{strides[1]} + 1 \\\\
            W_{out}&= \\frac{(W_{in} + 2 * paddings[2] - (dilations[2] * (W_f - 1) + 1))}{strides[2]} + 1

    Args:
        input (Variable): The input image with [N, C, D, H, W] format.
            num_filters(int): The number of filter. It is as same as the output
            image channel.
        filter_size (int|tuple|None): The filter size. If filter_size is a tuple,
1578
            it must contain three integers, (filter_size_D, filter_size_H, filter_size_W).
C
chengduoZH 已提交
1579 1580
            Otherwise, the filter will be a square.
        stride (int|tuple): The stride size. If stride is a tuple, it must
1581
            contain three integers, (stride_D, stride_H, stride_W). Otherwise, the
C
chengduoZH 已提交
1582 1583
            stride_D = stride_H = stride_W = stride. Default: stride = 1.
        padding (int|tuple): The padding size. If padding is a tuple, it must
1584
            contain three integers, (padding_D, padding_H, padding_W). Otherwise, the
C
chengduoZH 已提交
1585 1586
            padding_D = padding_H = padding_W = padding. Default: padding = 0.
        dilation (int|tuple): The dilation size. If dilation is a tuple, it must
1587
            contain three integers, (dilation_D, dilation_H, dilation_W). Otherwise, the
C
chengduoZH 已提交
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
            dilation_D = dilation_H = dilation_W = dilation. Default: dilation = 1.
        groups (int): The groups number of the Conv3d Layer. According to grouped
            convolution in Alex Krizhevsky's Deep CNN paper: when group=2,
            the first half of the filters is only connected to the first half
            of the input channels, while the second half of the filters is only
            connected to the second half of the input channels. Default: groups=1
        param_attr (ParamAttr): The parameters to the Conv3d Layer. Default: None
        bias_attr (ParamAttr): Bias parameter for the Conv3d layer. Default: None
        use_cudnn (bool): Use cudnn kernel or not, it is valid only when the cudnn
            library is installed. Default: True
        use_mkldnn (bool): Use mkldnn kernels or not.
        act (str): Activation type. Default: None
        name (str|None): A name for this layer(optional). If set None, the layer
            will be named automatically.

    Returns:
        Variable: The tensor variable storing the convolution and \
                  non-linearity activation result.

    Raises:
        ValueError: If the shapes of input, filter_size, stride, padding and
                    groups mismatch.

    Examples:
        .. code-block:: python

1614 1615
          data = fluid.layers.data(name='data', shape=[3, 12, 32, 32], dtype='float32')
          conv3d = fluid.layers.conv3d(input=data, num_filters=2, filter_size=3, act="relu")
C
chengduoZH 已提交
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
    """

    l_type = 'conv3d'

    helper = LayerHelper(l_type, **locals())
    dtype = helper.input_dtype()

    num_channels = input.shape[1]

    if groups is None:
        num_filter_channels = num_channels
    else:
        if num_channels % groups != 0:
            raise ValueError("num_channels must be divisible by groups.")
        num_filter_channels = num_channels / groups

    filter_size = utils.convert_to_list(filter_size, 3, 'filter_size')
    stride = utils.convert_to_list(stride, 3, 'stride')
    padding = utils.convert_to_list(padding, 3, 'padding')
    dilation = utils.convert_to_list(dilation, 3, 'dilation')

    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")

    input_shape = input.shape
    filter_shape = [num_filters, num_filter_channels] + filter_size

    def _get_default_param_initializer():
        std = (2.0 / (filter_size[0]**3 * num_channels))**0.5
        return Normal(0.0, std, 0)

    filter_param = helper.create_parameter(
        attr=helper.param_attr,
        shape=filter_shape,
        dtype=dtype,
        default_initializer=_get_default_param_initializer())

    pre_bias = helper.create_tmp_variable(dtype)

    helper.append_op(
        type=l_type,
        inputs={
            'Input': input,
            'Filter': filter_param,
        },
        outputs={"Output": pre_bias},
        attrs={
            'strides': stride,
            'paddings': padding,
            'dilations': dilation,
            'groups': groups,
            'use_cudnn': use_cudnn,
            'use_mkldnn': use_mkldnn
        })

1671
    pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)
C
chengduoZH 已提交
1672 1673 1674 1675

    return helper.append_activation(pre_act)


F
fengjiayi 已提交
1676
def sequence_pool(input, pool_type):
Y
Yu Yang 已提交
1677
    """
Y
yangyaming 已提交
1678 1679 1680
    This function add the operator for sequence pooling.
    It pools features of all time-steps of each instance, and is applied
    on top of the input using pool_type mentioned in the parameters.
L
Luo Tao 已提交
1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691

    It supports four pool_type:

    - average: :math:`Out[i] = \\frac{\sum_i X_i}{N}`
    - sum:     :math:`Out[i] = \sum_jX_{ij}`
    - sqrt:    :math:`Out[i] = \\frac{\sum_jX_{ij}}{\sqrt{len(X_i)}}`
    - max:     :math:`Out[i] = max(X_i)`

    .. code-block:: text

       x is a 1-level LoDTensor:
1692
         x.lod = [[2, 3, 2]]
L
Luo Tao 已提交
1693 1694 1695 1696 1697
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
1698
         with condition len(x.lod[-1]) == out.dims[0]
L
Luo Tao 已提交
1699 1700 1701 1702 1703 1704 1705

       for different pool_type:
         average: out.data = [2, 4, 3], where 2=(1+3)/2, 4=(2+4+6)/3, 3=(5+1)/2
         sum    : out.data = [4, 12, 6], where 4=1+3, 12=2+4+6, 6=5+1
         sqrt   : out.data = [2.82, 6.93, 4.24], where 2.82=(1+3)/sqrt(2),
                    6.93=(2+4+6)/sqrt(3), 4.24=(5+1)/sqrt(2)
         max    : out.data = [3, 6, 5], where 3=max(1,3), 6=max(2,4,6), 5=max(5,1)
1706 1707
         last   : out.data = [3, 6, 1], where 3=last(1,3), 6=last(2,4,6), 1=last(5,1)
         first  : out.data = [1, 2, 5], where 1=first(1,3), 2=first(2,4,6), 5=first(5,1)
F
fengjiayi 已提交
1708

L
Luo Tao 已提交
1709 1710
    Args:
        input(variable): The input variable which is a LoDTensor.
Y
yangyaming 已提交
1711
        pool_type (string): The pooling type of sequence_pool.
L
Luo Tao 已提交
1712 1713 1714 1715 1716 1717 1718 1719
            It supports average, sum, sqrt and max.

    Returns:
        The sequence pooling variable which is a Tensor.

    Examples:

        .. code-block:: python
F
fengjiayi 已提交
1720

Y
yangyaming 已提交
1721
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1722 1723 1724 1725 1726
                              dtype='float32', lod_level=1)
             avg_x = fluid.layers.sequence_pool(input=x, pool_type='average')
             sum_x = fluid.layers.sequence_pool(input=x, pool_type='sum')
             sqrt_x = fluid.layers.sequence_pool(input=x, pool_type='sqrt')
             max_x = fluid.layers.sequence_pool(input=x, pool_type='max')
1727 1728
             last_x = fluid.layers.sequence_pool(input=x, pool_type='last')
             first_x = fluid.layers.sequence_pool(input=x, pool_type='first')
Y
Yu Yang 已提交
1729
    """
F
fengjiayi 已提交
1730
    helper = LayerHelper('sequence_pool', **locals())
Y
Yu Yang 已提交
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
    dtype = helper.input_dtype()
    pool_out = helper.create_tmp_variable(dtype)
    max_index = helper.create_tmp_variable(dtype)

    helper.append_op(
        type="sequence_pool",
        inputs={"X": input},
        outputs={"Out": pool_out,
                 "MaxIndex": max_index},
        attrs={"pooltype": pool_type.upper()})

Y
yangyaming 已提交
1742 1743 1744 1745 1746
    # when pool_type is max, variable max_index is initialized,
    # so we stop the gradient explicitly here
    if pool_type == 'max':
        max_index.stop_gradient = True

Y
Yu Yang 已提交
1747 1748 1749
    return pool_out


F
fengjiayi 已提交
1750
def sequence_first_step(input):
L
Luo Tao 已提交
1751
    """
L
Luo Tao 已提交
1752
    This function gets the first step of sequence.
L
Luo Tao 已提交
1753 1754 1755 1756

    .. code-block:: text

       x is a 1-level LoDTensor:
1757
         x.lod = [[2, 3, 2]]
L
Luo Tao 已提交
1758 1759 1760 1761 1762
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
1763
         with condition len(x.lod[-1]) == out.dims[0]
L
Luo Tao 已提交
1764
         out.data = [1, 2, 5], where 1=first(1,3), 2=first(2,4,6), 5=first(5,1)
F
fengjiayi 已提交
1765

L
Luo Tao 已提交
1766 1767 1768 1769 1770 1771 1772 1773 1774
    Args:
        input(variable): The input variable which is a LoDTensor.

    Returns:
        The sequence's first step variable which is a Tensor.

    Examples:

        .. code-block:: python
F
fengjiayi 已提交
1775

Y
yangyaming 已提交
1776
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1777 1778 1779
                              dtype='float32', lod_level=1)
             x_first_step = fluid.layers.sequence_first_step(input=x)
    """
1780 1781 1782
    return sequence_pool(input=input, pool_type="first")


F
fengjiayi 已提交
1783
def sequence_last_step(input):
L
Luo Tao 已提交
1784
    """
L
Luo Tao 已提交
1785
    This function gets the last step of sequence.
L
Luo Tao 已提交
1786 1787 1788 1789

    .. code-block:: text

       x is a 1-level LoDTensor:
1790
         x.lod = [[2, 3, 2]]
L
Luo Tao 已提交
1791 1792 1793 1794 1795
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
1796
         with condition len(x.lod[-1]) == out.dims[0]
L
Luo Tao 已提交
1797
         out.data = [3, 6, 1], where 3=last(1,3), 6=last(2,4,6), 1=last(5,1)
F
fengjiayi 已提交
1798

L
Luo Tao 已提交
1799 1800 1801 1802 1803 1804 1805 1806 1807
    Args:
        input(variable): The input variable which is a LoDTensor.

    Returns:
        The sequence's last step variable which is a Tensor.

    Examples:

        .. code-block:: python
F
fengjiayi 已提交
1808

Y
yangyaming 已提交
1809
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1810 1811 1812
                              dtype='float32', lod_level=1)
             x_last_step = fluid.layers.sequence_last_step(input=x)
    """
1813 1814 1815
    return sequence_pool(input=input, pool_type="last")


F
fengjiayi 已提交
1816
@templatedoc()
Y
Yu Yang 已提交
1817
def pool2d(input,
C
chengduoZH 已提交
1818 1819
           pool_size=-1,
           pool_type="max",
C
chengduoZH 已提交
1820 1821
           pool_stride=1,
           pool_padding=0,
C
caoying03 已提交
1822
           global_pooling=False,
C
chengduoZH 已提交
1823
           use_cudnn=True,
1824
           ceil_mode=False,
1825
           use_mkldnn=False,
C
caoying03 已提交
1826
           name=None):
Y
Yu Yang 已提交
1827
    """
F
fengjiayi 已提交
1828
    ${comment}
1829 1830

    Args:
F
fengjiayi 已提交
1831
        input (Variable): The input tensor of pooling operator. The format of 
F
fengjiayi 已提交
1832 1833 1834
                          input tensor is NCHW, where N is batch size, C is 
                          the number of channels, H is the height of the 
                          feature, and W is the width of the feature.
F
fengjiayi 已提交
1835 1836
        pool_size (int): The side length of pooling windows. All pooling 
                         windows are squares with pool_size on a side.
F
fengjiayi 已提交
1837
        pool_type: ${pooling_type_comment}
1838 1839
        pool_stride (int): stride of the pooling layer.
        pool_padding (int): padding size.
F
fengjiayi 已提交
1840 1841 1842 1843
        global_pooling: ${global_pooling_comment}
        use_cudnn: ${use_cudnn_comment}
        ceil_mode: ${ceil_mode_comment}
        use_mkldnn: ${use_mkldnn_comment}
F
fengjiayi 已提交
1844 1845 1846
        name (str|None): A name for this layer(optional). If set None, the 
                        layer will be named automatically.

1847
    Returns:
F
fengjiayi 已提交
1848
        Variable: The pooling result.
F
fengjiayi 已提交
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866

    Raises:
        ValueError: If 'pool_type' is not "max" nor "avg"
        ValueError: If 'global_pooling' is False and 'pool_size' is -1
        ValueError: If 'use_cudnn' is not a bool value.

    Examples:

        .. code-block:: python

          data = fluid.layers.data(
              name='data', shape=[3, 32, 32], dtype='float32')
          conv2d = fluid.layers.pool2d(
                            input=data, 
                            pool_size=2, 
                            pool_type='max', 
                            pool_stride=1, 
                            global_pooling=False)
Y
Yu Yang 已提交
1867 1868 1869 1870 1871
    """
    if pool_type not in ["max", "avg"]:
        raise ValueError(
            "Unknown pool_type: '%s'. It can only be 'max' or 'avg'.",
            str(pool_type))
C
chengduoZH 已提交
1872

C
chengduoZH 已提交
1873 1874 1875 1876 1877
    if global_pooling is False and pool_size == -1:
        raise ValueError(
            "When the global_pooling is False, pool_size must be passed "
            "and be a valid value. Received pool_size: " + str(pool_size))

C
chengduoZH 已提交
1878 1879 1880 1881
    pool_size = utils.convert_to_list(pool_size, 2, 'pool_size')
    pool_padding = utils.convert_to_list(pool_padding, 2, 'pool_padding')
    pool_stride = utils.convert_to_list(pool_stride, 2, 'pool_stride')

C
chengduoZH 已提交
1882 1883
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
Y
Yu Yang 已提交
1884

C
Add doc  
chengduoZH 已提交
1885
    l_type = 'pool2d'
1886 1887

    helper = LayerHelper(l_type, **locals())
Y
Yu Yang 已提交
1888 1889 1890 1891
    dtype = helper.input_dtype()
    pool_out = helper.create_tmp_variable(dtype)

    helper.append_op(
1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
        type=l_type,
        inputs={"X": input},
        outputs={"Out": pool_out},
        attrs={
            "pooling_type": pool_type,
            "ksize": pool_size,
            "global_pooling": global_pooling,
            "strides": pool_stride,
            "paddings": pool_padding,
            "use_cudnn": use_cudnn,
            "ceil_mode": ceil_mode,
            "use_mkldnn": use_mkldnn
        })

    return pool_out


def pool3d(input,
           pool_size=-1,
           pool_type="max",
           pool_stride=1,
           pool_padding=0,
           global_pooling=False,
           use_cudnn=True,
           ceil_mode=False,
           use_mkldnn=False,
           name=None):
    """
    This function adds the operator for pooling in 3-dimensions, using the
Y
Yu Yang 已提交
1921
    pooling configurations mentioned in input parameters.
1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934

    Args:
        input (Variable): ${input_comment}
        pool_size (int): ${ksize_comment}
        pool_type (str): ${pooling_type_comment}
        pool_stride (int): stride of the pooling layer.
        pool_padding (int): padding size.
        global_pooling (bool): ${global_pooling_comment}
        use_cudnn (bool): ${use_cudnn_comment}
        ceil_mode (bool): ${ceil_mode_comment}
        use_mkldnn (bool): ${use_mkldnn_comment}
        name (str): A name for this layer(optional). If set None, the layer
            will be named automatically.
1935

1936
    Returns:
1937
        Variable: output of pool3d layer.
Y
Yu Yang 已提交
1938 1939 1940 1941 1942
    """
    if pool_type not in ["max", "avg"]:
        raise ValueError(
            "Unknown pool_type: '%s'. It can only be 'max' or 'avg'.",
            str(pool_type))
C
chengduoZH 已提交
1943

C
chengduoZH 已提交
1944 1945 1946 1947 1948
    if global_pooling is False and pool_size == -1:
        raise ValueError(
            "When the global_pooling is False, pool_size must be passed "
            "and be a valid value. Received pool_size: " + str(pool_size))

1949 1950 1951
    pool_size = utils.convert_to_list(pool_size, 3, 'pool_size')
    pool_padding = utils.convert_to_list(pool_padding, 3, 'pool_padding')
    pool_stride = utils.convert_to_list(pool_stride, 3, 'pool_stride')
C
chengduoZH 已提交
1952

C
chengduoZH 已提交
1953 1954
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
Y
Yu Yang 已提交
1955

1956 1957
    l_type = "pool3d"
    helper = LayerHelper(l_type, **locals())
Y
Yu Yang 已提交
1958 1959 1960 1961
    dtype = helper.input_dtype()
    pool_out = helper.create_tmp_variable(dtype)

    helper.append_op(
1962
        type=l_type,
Y
Yu Yang 已提交
1963 1964 1965 1966 1967 1968 1969
        inputs={"X": input},
        outputs={"Out": pool_out},
        attrs={
            "pooling_type": pool_type,
            "ksize": pool_size,
            "global_pooling": global_pooling,
            "strides": pool_stride,
C
chengduoZH 已提交
1970
            "paddings": pool_padding,
1971
            "use_cudnn": use_cudnn,
1972 1973
            "ceil_mode": ceil_mode,
            "use_mkldnn": use_mkldnn
Y
Yu Yang 已提交
1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
        })

    return pool_out


def batch_norm(input,
               act=None,
               is_test=False,
               momentum=0.9,
               epsilon=1e-05,
               param_attr=None,
               bias_attr=None,
C
caoying03 已提交
1986
               data_layout='NCHW',
Y
Yang Yang 已提交
1987
               in_place=False,
1988
               use_mkldnn=False,
1989 1990
               name=None,
               moving_mean_name=None,
W
wanghaoshuang 已提交
1991
               moving_variance_name=None,
W
wanghaoshuang 已提交
1992
               do_model_average_for_mean_and_var=False):
Y
Yu Yang 已提交
1993
    """
Q
qiaolongfei 已提交
1994 1995 1996 1997
    **Batch Normalization Layer**

    Can be used as a normalizer function for conv2d and fully_connected operations.
    The required data format for this layer is one of the following:
Q
qiaolongfei 已提交
1998

Q
qiaolongfei 已提交
1999
    1. NHWC `[batch, in_height, in_width, in_channels]`
Q
qiaolongfei 已提交
2000

Q
qiaolongfei 已提交
2001 2002
    2. NCHW `[batch, in_channels, in_height, in_width]`

Q
qiaolongfei 已提交
2003 2004 2005
    Refer to `Batch Normalization: Accelerating Deep Network Training by Reducing
    Internal Covariate Shift <https://arxiv.org/pdf/1502.03167.pdf>`_
    for more details.
Q
qiaolongfei 已提交
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

    :math:`input` is the input features over a mini-batch.

    ..  math::

        \\mu_{\\beta} &\\gets \\frac{1}{m} \\sum_{i=1}^{m} x_i \\qquad &//\\
        \ mini-batch\ mean \\\\
        \\sigma_{\\beta}^{2} &\\gets \\frac{1}{m} \\sum_{i=1}^{m}(x_i - \\
        \\mu_{\\beta})^2 \\qquad &//\ mini-batch\ variance \\\\
        \\hat{x_i} &\\gets \\frac{x_i - \\mu_\\beta} {\\sqrt{\\
        \\sigma_{\\beta}^{2} + \\epsilon}} \\qquad &//\ normalize \\\\
        y_i &\\gets \\gamma \\hat{x_i} + \\beta \\qquad &//\ scale\ and\ shift
2018 2019

    Args:
Q
qiaolongfei 已提交
2020
        input(variable): The input variable which is a LoDTensor.
Q
qiaolongfei 已提交
2021 2022 2023 2024
        act(string, Default None): Activation type, linear|relu|prelu|...
        is_test(bool, Default False): Used for training or training.
        momentum(float, Default 0.9):
        epsilon(float, Default 1e-05):
Q
qiaolongfei 已提交
2025 2026 2027
        param_attr(ParamAttr): The parameter attribute for Parameter `scale`.
        bias_attr(ParamAttr): The parameter attribute for Parameter `bias`.
        data_layout(string, default NCHW): NCHW|NHWC
Q
qiaolongfei 已提交
2028
        in_place(bool, Default False): Make the input and output of batch norm reuse memory.
Q
qiaolongfei 已提交
2029 2030 2031 2032 2033
        use_mkldnn(bool, Default false): ${use_mkldnn_comment}
        name(string, Default None): A name for this layer(optional). If set None, the layer
            will be named automatically.
        moving_mean_name(string, Default None): The name of moving_mean which store the global Mean.
        moving_variance_name(string, Default None): The name of the moving_variance which store the global Variance.
Q
qiaolongfei 已提交
2034
        do_model_average_for_mean_and_var(bool, Default False): Do model average for mean and variance or not.
2035 2036

    Returns:
Q
qiaolongfei 已提交
2037
        Variable: A tensor variable which is the result after applying batch normalization on the input.
Q
qiaolongfei 已提交
2038 2039 2040 2041 2042 2043 2044

    Examples:

        .. code-block:: python

            hidden1 = fluid.layers.fc(input=x, size=200, param_attr='fc1.w')
            hidden2 = fluid.layers.batch_norm(input=hidden1)
Y
Yu Yang 已提交
2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067
    """
    helper = LayerHelper('batch_norm', **locals())
    dtype = helper.input_dtype()

    input_shape = input.shape
    if data_layout == 'NCHW':
        channel_num = input_shape[1]
    else:
        if data_layout == 'NHWC':
            channel_num = input_shape[-1]
        else:
            raise ValueError("unsupported data layout:" + data_layout)

    param_shape = [channel_num]

    # create parameter
    scale = helper.create_parameter(
        attr=helper.param_attr,
        shape=param_shape,
        dtype=dtype,
        default_initializer=Constant(1.0))

    bias = helper.create_parameter(
2068
        attr=helper.bias_attr, shape=param_shape, dtype=dtype, is_bias=True)
Y
Yu Yang 已提交
2069

2070 2071
    mean = helper.create_parameter(
        attr=ParamAttr(
W
wanghaoshuang 已提交
2072 2073 2074
            name=moving_mean_name,
            initializer=Constant(0.0),
            trainable=False,
W
wanghaoshuang 已提交
2075
            do_model_average=do_model_average_for_mean_and_var),
Q
QI JUN 已提交
2076
        shape=param_shape,
2077 2078 2079 2080 2081 2082 2083
        dtype=input.dtype)
    mean.stop_gradient = True

    variance = helper.create_parameter(
        attr=ParamAttr(
            name=moving_variance_name,
            initializer=Constant(1.0),
W
wanghaoshuang 已提交
2084
            trainable=False,
W
wanghaoshuang 已提交
2085
            do_model_average=do_model_average_for_mean_and_var),
Q
QI JUN 已提交
2086
        shape=param_shape,
2087 2088
        dtype=input.dtype)
    variance.stop_gradient = True
Y
Yu Yang 已提交
2089 2090 2091 2092 2093 2094

    # create output
    # mean and mean_out share the same memory
    mean_out = mean
    # variance and variance out share the same memory
    variance_out = variance
Q
QI JUN 已提交
2095 2096
    saved_mean = helper.create_tmp_variable(dtype=dtype, stop_gradient=True)
    saved_variance = helper.create_tmp_variable(dtype=dtype, stop_gradient=True)
Y
Yu Yang 已提交
2097

Y
Yang Yang 已提交
2098
    batch_norm_out = input if in_place else helper.create_tmp_variable(dtype)
Y
Yu Yang 已提交
2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115

    helper.append_op(
        type="batch_norm",
        inputs={
            "X": input,
            "Scale": scale,
            "Bias": bias,
            "Mean": mean,
            "Variance": variance
        },
        outputs={
            "Y": batch_norm_out,
            "MeanOut": mean_out,
            "VarianceOut": variance_out,
            "SavedMean": saved_mean,
            "SavedVariance": saved_variance
        },
2116 2117 2118 2119 2120 2121
        attrs={
            "momentum": momentum,
            "epsilon": epsilon,
            "is_test": is_test,
            "use_mkldnn": use_mkldnn
        })
Y
Yu Yang 已提交
2122 2123 2124 2125

    return helper.append_activation(batch_norm_out)


Y
yuyang18 已提交
2126
@templatedoc()
G
guosheng 已提交
2127 2128 2129 2130 2131 2132 2133 2134 2135 2136
def layer_norm(input,
               scale=True,
               shift=True,
               begin_norm_axis=1,
               epsilon=1e-05,
               param_attr=None,
               bias_attr=None,
               act=None,
               name=None):
    """
Y
yuyang18 已提交
2137
    ${comment}
G
guosheng 已提交
2138 2139 2140

    The formula is as follows:

Y
yuyang18 已提交
2141
    ..  math::
G
guosheng 已提交
2142 2143 2144 2145 2146 2147 2148

        \\mu & = \\frac{1}{H}\\sum_{i=1}^{H} a_i

        \\sigma & = \\sqrt{\\frac{1}{H}\sum_{i=1}^{H}(a_i - \\mu)^2}

        h & = f(\\frac{g}{\\sigma}(a - \\mu) + b)

Y
yuyang18 已提交
2149 2150 2151 2152 2153 2154 2155 2156
    * :math:`a`: the vector representation of the summed inputs to the neurons
    in that layer.

    * :math:`H`: the number of hidden units in a layers

    * :math:`g`: the trainable scale parameter.

    * :math:`b`: the trainable bias parameter.
Y
yuyang18 已提交
2157

G
guosheng 已提交
2158 2159
    Args:
        input(Variable): The input tensor variable.
2160
        scale(bool): Whether to learn the adaptive gain :math:`g` after
G
guosheng 已提交
2161
            normalization.
2162
        shift(bool): Whether to learn the adaptive bias :math:`b` after
G
guosheng 已提交
2163
            normalization.
2164
        begin_norm_axis(bool): The normalization will be performed along
G
guosheng 已提交
2165
            dimensions from :attr:`begin_norm_axis` to :attr:`rank(input)`.
2166
        epsilon(float): The small value added to the variance to prevent
G
guosheng 已提交
2167 2168 2169 2170 2171 2172
            division by zero.
        param_attr(ParamAttr|None): The parameter attribute for the learnable
            gain :math:`g`.
        bias_attr(ParamAttr|None): The parameter attribute for the learnable
            bias :math:`b`.
        act(str): Activation to be applied to the output of layer normalizaiton.
2173
        name (str): The name of this layer. It is optional.
G
guosheng 已提交
2174 2175

    Returns:
Y
yuyang18 已提交
2176
        ${y_comment}
G
guosheng 已提交
2177 2178 2179

    Examples:

Y
yuyang18 已提交
2180 2181 2182
        >>> data = fluid.layers.data(name='data', shape=[3, 32, 32],
        >>>                          dtype='float32')
        >>> x = fluid.layers.layer_norm(input=data, begin_norm_axis=1)
G
guosheng 已提交
2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197
    """
    helper = LayerHelper('layer_norm', **locals())
    dtype = helper.input_dtype()

    # create intput and parameters
    inputs = {'X': input}
    input_shape = input.shape
    param_shape = [reduce(lambda x, y: x * y, input_shape[begin_norm_axis:])]
    if scale:
        scale = helper.create_parameter(
            attr=helper.param_attr,
            shape=param_shape,
            dtype=dtype,
            default_initializer=Constant(1.0))
        inputs['Scale'] = scale
G
guosheng 已提交
2198
    if shift:
G
guosheng 已提交
2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222
        assert bias_attr is not False
        bias = helper.create_parameter(
            attr=helper.bias_attr, shape=param_shape, dtype=dtype, is_bias=True)
        inputs['Bias'] = bias

    # create output
    mean_out = helper.create_tmp_variable(dtype=dtype, stop_gradient=True)
    variance_out = helper.create_tmp_variable(dtype=dtype, stop_gradient=True)
    layer_norm_out = helper.create_tmp_variable(dtype)

    helper.append_op(
        type="layer_norm",
        inputs=inputs,
        outputs={
            "Y": layer_norm_out,
            "Mean": mean_out,
            "Variance": variance_out,
        },
        attrs={"epsilon": epsilon,
               "begin_norm_axis": begin_norm_axis})

    return helper.append_activation(layer_norm_out)


C
caoying03 已提交
2223
def beam_search_decode(ids, scores, name=None):
2224
    """
D
dzhwinter 已提交
2225 2226
    Beam Search Decode

D
dzhwinter 已提交
2227 2228
    This layers is to pack the output of beam search layer into sentences and
    associated scores. It is usually called after the beam search layer.
D
dzhwinter 已提交
2229 2230 2231 2232 2233 2234
    Typically, the output of beam search layer is a tensor of selected ids, with
    a tensor of the score of each id. Beam search layer's output ids, however, 
    are generated directly during the tree search, and they are stacked by each 
    level of the search tree. Thus we need to reorganize them into sentences, 
    based on the score of each id. This layer takes the output of beam search
    layer as input and repack them into sentences.
2235 2236

    Args:
D
dzhwinter 已提交
2237 2238 2239
        ids (Variable): The selected ids, output of beam search layer. 
        scores (Variable): The associated scores of the ids, out put of beam
            search layer.
2240
        name (str): The name of this layer. It is optional.
F
fengjiayi 已提交
2241

2242
    Returns:
D
dzhwinter 已提交
2243 2244 2245 2246 2247
        tuple(Variable): a tuple of two output tensors: sentence_ids, sentence_scores.
        sentence_ids is a tensor with shape [size, length], where size is the
        beam size of beam search, and length is the length of each sentence. 
        Note that the length of sentences may vary.
        sentence_scores is a tensor with the same shape as sentence_ids.
D
dzhwinter 已提交
2248 2249 2250

    Examples:
        .. code-block:: python
D
dzhwinter 已提交
2251

D
dzhwinter 已提交
2252 2253 2254 2255
            ids, scores = fluid.layers.beam_search(
                pre_ids, ids, scores, beam_size, end_id)
            sentence_ids, sentence_scores = fluid.layers.beam_search_decode(
                ids, scores)
2256
    """
Y
Yu Yang 已提交
2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276
    helper = LayerHelper('beam_search_decode', **locals())
    sentence_ids = helper.create_tmp_variable(dtype=ids.dtype)
    sentence_scores = helper.create_tmp_variable(dtype=ids.dtype)

    helper.append_op(
        type="beam_search_decode",
        inputs={"Ids": ids,
                "Scores": scores},
        outputs={
            "SentenceIds": sentence_ids,
            "SentenceScores": sentence_scores
        })

    return sentence_ids, sentence_scores


def conv2d_transpose(input,
                     num_filters,
                     output_size=None,
                     filter_size=None,
C
chengduoZH 已提交
2277 2278 2279
                     padding=0,
                     stride=1,
                     dilation=1,
2280
                     groups=None,
C
caoying03 已提交
2281
                     param_attr=None,
2282
                     bias_attr=None,
C
chengduoZH 已提交
2283
                     use_cudnn=True,
2284
                     act=None,
C
caoying03 已提交
2285
                     name=None):
Y
Yu Yang 已提交
2286
    """
2287 2288 2289 2290 2291 2292 2293 2294
    **Convlution2D transpose layer**

    The convolution2D transpose layer calculates the output based on the input,
    filter, and dilations, strides, paddings. Input(Input) and output(Output)
    are in NCHW format. Where N is batch size, C is the number of channels,
    H is the height of the feature, and W is the width of the feature.
    Parameters(dilations, strides, paddings) are two elements. These two elements
    represent height and width, respectively. The details of convolution transpose
2295 2296
    layer, please refer to the following explanation and references
    `therein <http://www.matthewzeiler.com/wp-content/uploads/2017/07/cvpr2010.pdf>`_.
2297 2298 2299
    If bias attribution and activation type are provided, bias is added to
    the output of the convolution, and the corresponding activation function
    is applied to the final result.
2300 2301 2302 2303 2304

    For each input :math:`X`, the equation is:

    .. math::

2305
        Out = \sigma (W \\ast X + b)
2306

2307
    Where:
2308 2309 2310

    * :math:`X`: Input value, a tensor with NCHW format.
    * :math:`W`: Filter value, a tensor with MCHW format.
2311 2312 2313 2314
    * :math:`\\ast`: Convolution operation.
    * :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
    * :math:`\\sigma`: Activation function.
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
Y
Yu Yang 已提交
2315

2316 2317 2318 2319
    Example:

        - Input:

2320
          Input shape: :math:`(N, C_{in}, H_{in}, W_{in})`
2321

2322
          Filter shape: :math:`(C_{in}, C_{out}, H_f, W_f)`
2323 2324 2325

        - Output:

2326
          Output shape: :math:`(N, C_{out}, H_{out}, W_{out})`
2327 2328

        Where
Y
Yu Yang 已提交
2329

2330 2331 2332 2333
        .. math::

           H_{out} &= (H_{in} - 1) * strides[0] - 2 * paddings[0] + dilations[0] * (H_f - 1) + 1 \\\\
           W_{out} &= (W_{in} - 1) * strides[1] - 2 * paddings[1] + dilations[1] * (W_f - 1) + 1
Y
Yu Yang 已提交
2334 2335

    Args:
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368
        input(Variable): The input image with [N, C, H, W] format.
        num_filters(int): The number of the filter. It is as same as the output
            image channel.
        output_size(int|tuple|None): The output image size. If output size is a
            tuple, it must contain two integers, (image_H, image_W). This
            parameter only works when filter_size is None.
        filter_size(int|tuple|None): The filter size. If filter_size is a tuple,
            it must contain two integers, (filter_size_H, filter_size_W).
            Otherwise, the filter will be a square. None if use output size to
            calculate filter_size.
        padding(int|tuple): The padding size. If padding is a tuple, it must
            contain two integers, (padding_H, padding_W). Otherwise, the
            padding_H = padding_W = padding. Default: padding = 0.
        stride(int|tuple): The stride size. If stride is a tuple, it must
            contain two integers, (stride_H, stride_W). Otherwise, the
            stride_H = stride_W = stride. Default: stride = 1.
        dilation(int|tuple): The dilation size. If dilation is a tuple, it must
            contain two integers, (dilation_H, dilation_W). Otherwise, the
            dilation_H = dilation_W = dilation. Default: dilation = 1.
        groups(int): The groups number of the Conv2d transpose layer. Inspired by
            grouped convolution in Alex Krizhevsky's Deep CNN paper, in which
            when group=2, the first half of the filters is only connected to the
            first half of the input channels, while the second half of the
            filters is only connected to the second half of the input channels.
            Default: groups=1
        param_attr(ParamAttr): The parameters to the Conv2d_transpose Layer.
                               Default: None
        bias_attr(ParamAttr): Bias parameter for the Conv2d layer. Default: None
        use_cudnn(bool): Use cudnn kernel or not, it is valid only when the cudnn
            library is installed. Default: True
        act(str): Activation type. Default: None
        name(str|None): A name for this layer(optional). If set None, the layer
            will be named automatically.
Y
Yu Yang 已提交
2369 2370

    Returns:
2371
        Variable: The tensor variable storing the convolution transpose result.
2372 2373

    Raises:
2374 2375
        ValueError: If the shapes of input, filter_size, stride, padding and
                    groups mismatch.
2376 2377 2378 2379

    Examples:
       .. code-block:: python

2380 2381
          data = fluid.layers.data(name='data', shape=[3, 32, 32], dtype='float32')
          conv2d_transpose = fluid.layers.conv2d_transpose(input=data, num_filters=2, filter_size=3)
Y
Yu Yang 已提交
2382 2383 2384 2385 2386 2387
    """
    helper = LayerHelper("conv2d_transpose", **locals())
    if not isinstance(input, Variable):
        raise TypeError("Input of conv2d_transpose must be Variable")
    input_channel = input.shape[1]

C
chengduoZH 已提交
2388 2389 2390
    padding = utils.convert_to_list(padding, 2, 'padding')
    stride = utils.convert_to_list(stride, 2, 'stride')
    dilation = utils.convert_to_list(dilation, 2, 'dilation')
G
guosheng 已提交
2391

C
chengduoZH 已提交
2392 2393
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
G
guosheng 已提交
2394

Y
Yu Yang 已提交
2395 2396 2397 2398 2399
    if filter_size is None:
        if output_size is None:
            raise ValueError("output_size must be set when filter_size is None")
        if isinstance(output_size, int):
            output_size = [output_size, output_size]
G
guosheng 已提交
2400

Y
Yu Yang 已提交
2401 2402
        h_in = input.shape[2]
        w_in = input.shape[3]
G
guosheng 已提交
2403

C
chengduoZH 已提交
2404 2405 2406 2407
        filter_size_h = (output_size[0] - (h_in - 1) * stride[0] + 2 *
                         padding[0] - 1) / dilation[0] + 1
        filter_size_w = (output_size[1] - (w_in - 1) * stride[1] + 2 *
                         padding[1] - 1) / dilation[1] + 1
Y
Yu Yang 已提交
2408
        filter_size = [filter_size_h, filter_size_w]
C
chengduoZH 已提交
2409 2410 2411
    else:
        filter_size = utils.convert_to_list(filter_size, 2,
                                            'conv2d_transpose.filter_size')
2412

2413 2414
    groups = 1 if groups is None else groups
    filter_shape = [input_channel, num_filters / groups] + filter_size
Y
Yu Yang 已提交
2415 2416 2417
    img_filter = helper.create_parameter(
        dtype=input.dtype, shape=filter_shape, attr=helper.param_attr)

2418
    pre_bias = helper.create_tmp_variable(dtype=input.dtype)
Y
Yu Yang 已提交
2419 2420 2421 2422
    helper.append_op(
        type='conv2d_transpose',
        inputs={'Input': [input],
                'Filter': [img_filter]},
2423
        outputs={'Output': pre_bias},
C
chengduoZH 已提交
2424
        attrs={
2425 2426 2427 2428 2429
            'strides': stride,
            'paddings': padding,
            'dilations': dilation,
            'groups': groups,
            'use_cudnn': use_cudnn
Y
Yu Yang 已提交
2430 2431
        })

2432 2433 2434
    pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)
    out = helper.append_activation(pre_act)
    return out
Y
Yu Yang 已提交
2435 2436


2437
def conv3d_transpose(input,
Y
Yu Yang 已提交
2438 2439 2440
                     num_filters,
                     output_size=None,
                     filter_size=None,
C
chengduoZH 已提交
2441 2442 2443
                     padding=0,
                     stride=1,
                     dilation=1,
2444
                     groups=None,
C
caoying03 已提交
2445
                     param_attr=None,
2446
                     bias_attr=None,
C
chengduoZH 已提交
2447
                     use_cudnn=True,
2448
                     act=None,
C
caoying03 已提交
2449
                     name=None):
Y
Yu Yang 已提交
2450
    """
2451
    **Convlution3D transpose layer**
2452

2453
    The convolution3D transpose layer calculates the output based on the input,
2454
    filter, and dilations, strides, paddings. Input(Input) and output(Output)
2455 2456 2457 2458 2459 2460
    are in NCDHW format. Where N is batch size, C is the number of channels,
    D is the depth of the feature, H is the height of the feature, and W
    is the width of the feature. Parameters(dilations, strides, paddings) are
    two elements. These two elements represent height and width, respectively.
    The details of convolution transpose layer, please refer to the following
    explanation and references `therein <http://www.matthewzeiler.com/wp-content/uploads/2017/07/cvpr2010.pdf>`_.
2461 2462 2463
    If bias attribution and activation type are provided, bias is added to
    the output of the convolution, and the corresponding activation function
    is applied to the final result.
2464 2465 2466 2467 2468

    For each input :math:`X`, the equation is:

    .. math::

2469
        Out = \sigma (W \\ast X + b)
2470 2471 2472

    In the above equation:

2473 2474
    * :math:`X`: Input value, a tensor with NCDHW format.
    * :math:`W`: Filter value, a tensor with MCDHW format.
2475 2476 2477 2478
    * :math:`\\ast`: Convolution operation.
    * :math:`b`: Bias value, a 2-D tensor with shape [M, 1].
    * :math:`\\sigma`: Activation function.
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be different.
Y
Yu Yang 已提交
2479

2480 2481 2482 2483
    Example:

        - Input:

2484
          Input shape: :math:`(N, C_{in}, D_{in}, H_{in}, W_{in})`
2485

2486
          Filter shape: :math:`(C_{in}, C_{out}, D_f, H_f, W_f)`
2487 2488 2489

        - Output:

2490
          Output shape: :math:`(N, C_{out}, D_{out}, H_{out}, W_{out})`
2491 2492

        Where
Y
Yu Yang 已提交
2493

2494 2495
        .. math::

2496 2497 2498
           D_{out} &= (D_{in} - 1) * strides[0] - 2 * paddings[0] + dilations[0] * (D_f - 1) + 1 \\\\
           H_{out} &= (H_{in} - 1) * strides[1] - 2 * paddings[1] + dilations[1] * (H_f - 1) + 1 \\\\
           W_{out} &= (W_{in} - 1) * strides[2] - 2 * paddings[2] + dilations[2] * (W_f - 1) + 1
Y
Yu Yang 已提交
2499 2500

    Args:
2501
        input(Variable): The input image with [N, C, D, H, W] format.
2502 2503 2504
        num_filters(int): The number of the filter. It is as same as the output
            image channel.
        output_size(int|tuple|None): The output image size. If output size is a
2505
            tuple, it must contain three integers, (image_D, image_H, image_W). This
2506 2507
            parameter only works when filter_size is None.
        filter_size(int|tuple|None): The filter size. If filter_size is a tuple,
2508
            it must contain three integers, (filter_size_D, filter_size_H, filter_size_W).
2509 2510 2511
            Otherwise, the filter will be a square. None if use output size to
            calculate filter_size.
        padding(int|tuple): The padding size. If padding is a tuple, it must
2512 2513
            contain three integers, (padding_D, padding_H, padding_W). Otherwise, the
            padding_D = padding_H = padding_W = padding. Default: padding = 0.
2514
        stride(int|tuple): The stride size. If stride is a tuple, it must
2515 2516
            contain three integers, (stride_D, stride_H, stride_W). Otherwise, the
            stride_D = stride_H = stride_W = stride. Default: stride = 1.
2517
        dilation(int|tuple): The dilation size. If dilation is a tuple, it must
2518 2519 2520
            contain three integers, (dilation_D, dilation_H, dilation_W). Otherwise, the
            dilation_D = dilation_H = dilation_W = dilation. Default: dilation = 1.
        groups(int): The groups number of the Conv3d transpose layer. Inspired by
2521 2522 2523 2524 2525
            grouped convolution in Alex Krizhevsky's Deep CNN paper, in which
            when group=2, the first half of the filters is only connected to the
            first half of the input channels, while the second half of the
            filters is only connected to the second half of the input channels.
            Default: groups=1
2526 2527 2528
        param_attr(ParamAttr): The parameters to the Conv3d_transpose Layer.
            Default: None
        bias_attr(ParamAttr): Bias parameter for the Conv3d layer. Default: None
2529 2530 2531 2532 2533
        use_cudnn(bool): Use cudnn kernel or not, it is valid only when the cudnn
            library is installed. Default: True
        act(str): Activation type. Default: None
        name(str|None): A name for this layer(optional). If set None, the layer
            will be named automatically.
Y
Yu Yang 已提交
2534 2535

    Returns:
2536
        Variable: The tensor variable storing the convolution transpose result.
2537 2538

    Raises:
2539 2540
        ValueError: If the shapes of input, filter_size, stride, padding and
                    groups mismatch.
2541 2542 2543 2544

    Examples:
       .. code-block:: python

2545 2546
          data = fluid.layers.data(name='data', shape=[3, 12, 32, 32], dtype='float32')
          conv3d_transpose = fluid.layers.conv3d_transpose(input=data, num_filters=2, filter_size=3)
Y
Yu Yang 已提交
2547
    """
2548 2549
    l_type = "conv3d_transpose"
    helper = LayerHelper(l_type, **locals())
Y
Yu Yang 已提交
2550
    if not isinstance(input, Variable):
2551
        raise TypeError("Input of conv3d_transpose must be Variable")
Y
Yu Yang 已提交
2552 2553
    input_channel = input.shape[1]

2554 2555 2556
    padding = utils.convert_to_list(padding, 3, 'padding')
    stride = utils.convert_to_list(stride, 3, 'stride')
    dilation = utils.convert_to_list(dilation, 3, 'dilation')
C
chengduoZH 已提交
2557

C
chengduoZH 已提交
2558 2559 2560
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")

Y
Yu Yang 已提交
2561 2562 2563 2564 2565 2566
    if filter_size is None:
        if output_size is None:
            raise ValueError("output_size must be set when filter_size is None")
        if isinstance(output_size, int):
            output_size = [output_size, output_size]

2567 2568 2569
        d_in = input.shape[2]
        h_in = input.shape[3]
        w_in = input.shape[4]
C
chengduoZH 已提交
2570

2571
        filter_size_d = (output_size[0] - (d_in - 1) * stride[0] + 2 *
C
chengduoZH 已提交
2572
                         padding[0] - 1) / dilation[0] + 1
2573
        filter_size_h = (output_size[1] - (h_in - 1) * stride[1] + 2 *
C
chengduoZH 已提交
2574
                         padding[1] - 1) / dilation[1] + 1
2575 2576 2577
        filter_size_w = (output_size[2] - (w_in - 1) * stride[2] + 2 *
                         padding[2] - 1) / dilation[2] + 1
        filter_size = [filter_size_d, filter_size_h, filter_size_w]
C
chengduoZH 已提交
2578
    else:
2579 2580
        filter_size = utils.convert_to_list(filter_size, 3,
                                            'conv3d_transpose.filter_size')
Y
Yu Yang 已提交
2581

2582 2583
    groups = 1 if groups is None else groups
    filter_shape = [input_channel, num_filters / groups] + filter_size
Y
Yu Yang 已提交
2584 2585 2586
    img_filter = helper.create_parameter(
        dtype=input.dtype, shape=filter_shape, attr=helper.param_attr)

2587
    pre_bias = helper.create_tmp_variable(dtype=input.dtype)
Y
Yu Yang 已提交
2588
    helper.append_op(
2589
        type=l_type,
Y
Yu Yang 已提交
2590 2591
        inputs={'Input': [input],
                'Filter': [img_filter]},
2592
        outputs={'Output': pre_bias},
C
chengduoZH 已提交
2593 2594 2595 2596
        attrs={
            'strides': stride,
            'paddings': padding,
            'dilations': dilation,
2597
            'groups': groups,
C
chengduoZH 已提交
2598 2599
            'use_cudnn': use_cudnn
        })
Y
Yu Yang 已提交
2600

2601 2602
    pre_act = helper.append_bias_op(pre_bias, dim_start=1, dim_end=2)
    out = helper.append_activation(pre_act)
Y
Yu Yang 已提交
2603
    return out
Y
yangyaming 已提交
2604 2605


Y
yangyaming 已提交
2606
def sequence_expand(x, y, ref_level=-1, name=None):
2607
    """Sequence Expand Layer. This layer will expand the input variable **x**
Y
yangyaming 已提交
2608 2609 2610 2611
    according to specified level lod of **y**. Please note that lod level of
    **x** is at most 1 and rank of **x** is at least 2. When rank of **x**
    is greater than 2, then it would be viewed as a 2-D tensor.
    Following examples will explain how sequence_expand works:
2612 2613 2614 2615 2616

    .. code-block:: text

        * Case 1
            x is a LoDTensor:
2617
                x.lod  = [[2,        2]]
Y
yangyaming 已提交
2618
                x.data = [[a], [b], [c], [d]]
2619 2620 2621
                x.dims = [4, 1]

            y is a LoDTensor:
2622 2623
                y.lod = [[2,    2],
                         [3, 3, 1, 1]]
2624

Y
yangyaming 已提交
2625
            ref_level: 0
2626

Y
yangyaming 已提交
2627
            then output is a 1-level LoDTensor:
2628
                out.lod =  [[2,        2,        2,        2]]
Y
yangyaming 已提交
2629
                out.data = [[a], [b], [a], [b], [c], [d], [c], [d]]
2630 2631 2632 2633
                out.dims = [8, 1]

        * Case 2
            x is a Tensor:
Y
yangyaming 已提交
2634
                x.data = [[a], [b], [c]]
2635 2636 2637
                x.dims = [3, 1]

            y is a LoDTensor:
2638
                y.lod = [[2, 0, 3]]
2639

Y
yangyaming 已提交
2640
            ref_level: -1
2641

Y
yangyaming 已提交
2642 2643 2644
            then output is a Tensor:
                out.data = [[a], [a], [c], [c], [c]]
                out.dims = [5, 1]
2645 2646 2647
    Args:
        x (Variable): The input variable which is a Tensor or LoDTensor.
        y (Variable): The input variable which is a LoDTensor.
Y
yangyaming 已提交
2648 2649
        ref_level (int): Lod level of `y` to be referred by `x`. If set to -1,
                         refer the last level of lod.
C
caoying03 已提交
2650
        name(str|None): A name for this layer(optional). If set None, the layer
Y
yangyaming 已提交
2651
                        will be named automatically.
2652 2653 2654 2655 2656 2657 2658 2659 2660 2661

    Returns:
        Variable: The expanded variable which is a LoDTensor.

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[10], dtype='float32')
            y = fluid.layers.data(name='y', shape=[10, 20],
                             dtype='float32', lod_level=1)
Y
yangyaming 已提交
2662
            out = layers.sequence_expand(x=x, y=y, ref_level=0)
2663
    """
Y
yangyaming 已提交
2664
    helper = LayerHelper('sequence_expand', input=x, **locals())
2665 2666 2667
    dtype = helper.input_dtype()
    tmp = helper.create_tmp_variable(dtype)
    helper.append_op(
Y
yangyaming 已提交
2668 2669 2670 2671 2672
        type='sequence_expand',
        inputs={'X': x,
                'Y': y},
        outputs={'Out': tmp},
        attrs={'ref_level': ref_level})
2673
    return tmp
2674 2675


Q
Qiao Longfei 已提交
2676 2677 2678
def beam_search(pre_ids, ids, scores, beam_size, end_id, level=0):
    '''
    This function implements the beam search algorithm.
2679 2680 2681 2682 2683 2684 2685 2686

    Args:
        pre_ids (Variable): ${pre_ids_comment}
        ids (Variable): ${ids_comment}
        scores (Variable): ${scores_comment}
        beam_size (int): ${beam_size_comment}
        end_id (int): ${end_id_comment}
        level (int): ${level_comment}
F
fengjiayi 已提交
2687

2688 2689
    Returns:
        tuple: a tuple of beam_search output variables: selected_ids, selected_scores
Q
Qiao Longfei 已提交
2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
    '''
    helper = LayerHelper('beam_search', **locals())
    score_type = scores.dtype
    id_type = ids.dtype

    selected_scores = helper.create_tmp_variable(dtype=score_type)
    selected_ids = helper.create_tmp_variable(dtype=id_type)

    helper.append_op(
        type='beam_search',
        inputs={
            'pre_ids': pre_ids,
            'ids': ids,
            'scores': scores,
        },
        outputs={
            'selected_ids': selected_ids,
            'selected_scores': selected_scores,
        },
        attrs={
            # TODO(ChunweiYan) to assure other value support
            'level': level,
            'beam_size': beam_size,
            'end_id': end_id,
        })

    return selected_ids, selected_scores


Y
yangyaming 已提交
2719 2720 2721 2722
def lstm_unit(x_t,
              hidden_t_prev,
              cell_t_prev,
              forget_bias=0.0,
Y
yangyaming 已提交
2723
              param_attr=None,
C
caoying03 已提交
2724 2725
              bias_attr=None,
              name=None):
Y
yangyaming 已提交
2726 2727 2728 2729
    """Lstm unit layer. The equation of a lstm step is:

        .. math::

2730
            i_t & = \sigma(W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i)
Y
yangyaming 已提交
2731

2732
            f_t & = \sigma(W_{x_f}x_{t} + W_{h_f}h_{t-1} + b_f)
Y
yangyaming 已提交
2733

2734
            c_t & = f_tc_{t-1} + i_t tanh (W_{x_c}x_t + W_{h_c}h_{t-1} + b_c)
Y
yangyaming 已提交
2735

2736
            o_t & = \sigma(W_{x_o}x_{t} + W_{h_o}h_{t-1} + b_o)
Y
yangyaming 已提交
2737 2738 2739

            h_t & = o_t tanh(c_t)

2740 2741 2742 2743 2744 2745
    The inputs of lstm unit include :math:`x_t`, :math:`h_{t-1}` and
    :math:`c_{t-1}`. The 2nd dimensions of :math:`h_{t-1}` and :math:`c_{t-1}`
    should be same. The implementation separates the linear transformation and
    non-linear transformation apart. Here, we take :math:`i_t` as an example.
    The linear transformation is applied by calling a `fc` layer and the
    equation is:
Y
yangyaming 已提交
2746 2747 2748

        .. math::

2749
            L_{i_t} = W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i
Y
yangyaming 已提交
2750 2751 2752 2753 2754 2755 2756 2757

    The non-linear transformation is applied by calling `lstm_unit_op` and the
    equation is:

        .. math::

            i_t = \sigma(L_{i_t})

Y
yangyaming 已提交
2758
    This layer has two outputs including :math:`h_t` and :math:`o_t`.
Y
yangyaming 已提交
2759 2760

    Args:
Y
yangyaming 已提交
2761 2762 2763 2764 2765 2766
        x_t (Variable): The input value of current step, a 2-D tensor with shape
            M x N, M for batch size and N for input size.
        hidden_t_prev (Variable): The hidden value of lstm unit, a 2-D tensor
            with shape M x S, M for batch size and S for size of lstm unit.
        cell_t_prev (Variable): The cell value of lstm unit, a 2-D tensor with
            shape M x S, M for batch size and S for size of lstm unit.
Y
yangyaming 已提交
2767
        forget_bias (float): The forget bias of lstm unit.
Y
yangyaming 已提交
2768 2769
        param_attr (ParamAttr): The attributes of parameter weights, used to set
            initializer, name etc.
Y
yangyaming 已提交
2770 2771
        bias_attr (ParamAttr): The attributes of bias weights, if not False,
            bias weights will be created and be set to default value.
C
caoying03 已提交
2772 2773
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
Y
yangyaming 已提交
2774 2775

    Returns:
Y
yangyaming 已提交
2776
        tuple: The hidden value and cell value of lstm unit.
Y
yangyaming 已提交
2777 2778

    Raises:
2779 2780 2781 2782
        ValueError: The ranks of **x_t**, **hidden_t_prev** and **cell_t_prev**
                    not be 2 or the 1st dimensions of **x_t**, **hidden_t_prev**
                    and **cell_t_prev** not be the same or the 2nd dimensions of
                    **hidden_t_prev** and **cell_t_prev** not be the same.
Y
yangyaming 已提交
2783 2784 2785 2786 2787 2788

    Examples:

        .. code-block:: python

             x_t = fluid.layers.fc(input=x_t_data, size=10)
2789
             prev_hidden = fluid.layers.fc(input=prev_hidden_data, size=30)
Y
yangyaming 已提交
2790
             prev_cell = fluid.layers.fc(input=prev_cell_data, size=30)
Y
yangyaming 已提交
2791
             hidden_value, cell_value = fluid.layers.lstm_unit(x_t=x_t,
Y
yangyaming 已提交
2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807
                                                    hidden_t_prev=prev_hidden,
                                                    cell_t_prev=prev_cell)
    """
    helper = LayerHelper('lstm_unit', **locals())

    if len(x_t.shape) != 2:
        raise ValueError("Rank of x_t must be 2.")

    if len(hidden_t_prev.shape) != 2:
        raise ValueError("Rank of hidden_t_prev must be 2.")

    if len(cell_t_prev.shape) != 2:
        raise ValueError("Rank of cell_t_prev must be 2.")

    if x_t.shape[0] != hidden_t_prev.shape[0] or x_t.shape[
            0] != cell_t_prev.shape[0]:
Y
yangyaming 已提交
2808
        raise ValueError("The 1st dimensions of x_t, hidden_t_prev and "
2809 2810 2811 2812
                         "cell_t_prev must be the same.")

    if hidden_t_prev.shape[1] != cell_t_prev.shape[1]:
        raise ValueError("The 2nd dimensions of hidden_t_prev and "
Y
yangyaming 已提交
2813 2814
                         "cell_t_prev must be the same.")

Y
yangyaming 已提交
2815 2816 2817
    if bias_attr is None:
        bias_attr = ParamAttr()

Y
yangyaming 已提交
2818
    size = cell_t_prev.shape[1]
2819
    concat_out = concat(input=[x_t, hidden_t_prev], axis=1)
Y
yangyaming 已提交
2820 2821
    fc_out = fc(input=concat_out,
                size=4 * size,
Y
yangyaming 已提交
2822
                param_attr=param_attr,
2823
                bias_attr=bias_attr)
Y
yangyaming 已提交
2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835
    dtype = x_t.dtype
    c = helper.create_tmp_variable(dtype)
    h = helper.create_tmp_variable(dtype)

    helper.append_op(
        type='lstm_unit',
        inputs={"X": fc_out,
                "C_prev": cell_t_prev},
        outputs={"C": c,
                 "H": h},
        attrs={"forget_bias": forget_bias})

Y
yangyaming 已提交
2836
    return h, c
G
guosheng 已提交
2837 2838


C
caoying03 已提交
2839
def reduce_sum(input, dim=None, keep_dim=False, name=None):
G
guosheng 已提交
2840
    """
Y
yangyaming 已提交
2841
    Computes the sum of tensor elements over the given dimension.
G
guosheng 已提交
2842 2843 2844

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2845
        dim (list|int|None): The dimensions along which the sum is performed. If
Y
yangyaming 已提交
2846 2847
            :attr:`None`, sum all elements of :attr:`input` and return a
            Tensor variable with a single element, otherwise must be in the
W
whs 已提交
2848 2849
            range :math:`[-rank(input), rank(input))`. If :math:`dim[i] < 0`,
            the dimension to reduce is :math:`rank + dim[i]`.
2850
        keep_dim (bool|False): Whether to reserve the reduced dimension in the
Y
yangyaming 已提交
2851
            output Tensor. The result tensor will have one fewer dimension
G
guosheng 已提交
2852
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2853 2854
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
G
guosheng 已提交
2855 2856 2857

    Returns:
        Variable: The reduced Tensor variable.
F
fengjiayi 已提交
2858

G
guosheng 已提交
2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
    Examples:
        .. code-block:: python

            # x is a Tensor variable with following elements:
            #    [[0.2, 0.3, 0.5, 0.9]
            #     [0.1, 0.2, 0.6, 0.7]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_sum(x)  # [3.5]
            fluid.layers.reduce_sum(x, dim=0)  # [0.3, 0.5, 1.1, 1.6]
            fluid.layers.reduce_sum(x, dim=-1)  # [1.9, 1.6]
            fluid.layers.reduce_sum(x, dim=1, keep_dim=True)  # [[1.9], [1.6]]
W
whs 已提交
2870 2871 2872 2873 2874 2875 2876 2877

            # x is a Tensor variable with shape [2, 2, 2] and elements as below:
            #      [[[1, 2], [3, 4]],
            #      [[5, 6], [7, 8]]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_sum(x, dim=[1, 2]) # [10, 26]
            fluid.layers.reduce_sum(x, dim=[0, 1]) # [16, 20]

G
guosheng 已提交
2878 2879 2880
    """
    helper = LayerHelper('reduce_sum', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2881 2882
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
G
guosheng 已提交
2883 2884 2885 2886 2887
    helper.append_op(
        type='reduce_sum',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2888
            'dim': dim if dim != None else [0],
G
guosheng 已提交
2889 2890 2891 2892
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
G
guosheng 已提交
2893 2894


C
caoying03 已提交
2895
def reduce_mean(input, dim=None, keep_dim=False, name=None):
G
guosheng 已提交
2896
    """
Y
Yibing Liu 已提交
2897
    Computes the mean of the input tensor's elements along the given dimension.
G
guosheng 已提交
2898 2899 2900

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
Y
Yibing Liu 已提交
2901 2902 2903
        dim (list|int|None): The dimension along which the mean is computed. If
            `None`, compute the mean over all elements of :attr:`input`
            and return a variable with a single element, otherwise it
Y
yangyaming 已提交
2904
            must be in the range :math:`[-rank(input), rank(input))`. If
Y
Yibing Liu 已提交
2905 2906
            :math:`dim[i] < 0`, the dimension to reduce is 
            :math:`rank(input) + dim[i]`.
Y
yangyaming 已提交
2907 2908
        keep_dim (bool): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
G
guosheng 已提交
2909
            than the :attr:`input` unless :attr:`keep_dim` is true.
Y
Yibing Liu 已提交
2910
        name(str|None): A name for this layer(optional). If set `None`, the layer
C
caoying03 已提交
2911
                       will be named automatically.
G
guosheng 已提交
2912 2913

    Returns:
Y
Yibing Liu 已提交
2914
        Variable: The reduced mean Variable.
F
fengjiayi 已提交
2915

G
guosheng 已提交
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925
    Examples:
        .. code-block:: python

            # x is a Tensor variable with following elements:
            #    [[0.2, 0.3, 0.5, 0.9]
            #     [0.1, 0.2, 0.6, 0.7]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_mean(x)  # [0.4375]
            fluid.layers.reduce_mean(x, dim=0)  # [0.15, 0.25, 0.55, 0.8]
            fluid.layers.reduce_mean(x, dim=-1)  # [0.475, 0.4]
F
stash  
fengjiayi 已提交
2926 2927
            fluid.layers.reduce_mean(
                x, dim=1, keep_dim=True)  # [[0.475], [0.4]]
W
whs 已提交
2928 2929 2930 2931 2932 2933 2934

            # x is a Tensor variable with shape [2, 2, 2] and elements as below:
            #      [[[1.0, 2.0], [3.0, 4.0]],
            #      [[5.0, 6.0], [7.0, 8.0]]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_mean(x, dim=[1, 2]) # [2.5, 6.5]
            fluid.layers.reduce_mean(x, dim=[0, 1]) # [4.0, 5.0]
G
guosheng 已提交
2935 2936 2937
    """
    helper = LayerHelper('reduce_mean', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2938 2939
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
G
guosheng 已提交
2940 2941 2942 2943 2944
    helper.append_op(
        type='reduce_mean',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2945
            'dim': dim if dim != None else [0],
G
guosheng 已提交
2946 2947 2948 2949
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
2950 2951


C
caoying03 已提交
2952
def reduce_max(input, dim=None, keep_dim=False, name=None):
2953
    """
Y
yangyaming 已提交
2954
    Computes the maximum of tensor elements over the given dimension.
2955 2956 2957

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2958
        dim (list|int|None): The dimension along which the maximum is computed.
Y
yangyaming 已提交
2959 2960 2961
            If :attr:`None`, compute the maximum over all elements of
            :attr:`input` and return a Tensor variable with a single element,
            otherwise must be in the range :math:`[-rank(input), rank(input))`.
W
whs 已提交
2962
            If :math:`dim[i] < 0`, the dimension to reduce is :math:`rank + dim[i]`.
Y
yangyaming 已提交
2963 2964
        keep_dim (bool): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
2965
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2966 2967
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
2968 2969 2970

    Returns:
        Variable: The reduced Tensor variable.
Y
yangyaming 已提交
2971

2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
    Examples:
        .. code-block:: python

            # x is a Tensor variable with following elements:
            #    [[0.2, 0.3, 0.5, 0.9]
            #     [0.1, 0.2, 0.6, 0.7]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_max(x)  # [0.9]
            fluid.layers.reduce_max(x, dim=0)  # [0.2, 0.3, 0.6, 0.9]
            fluid.layers.reduce_max(x, dim=-1)  # [0.9, 0.7]
            fluid.layers.reduce_max(x, dim=1, keep_dim=True)  # [[0.9], [0.7]]
W
whs 已提交
2983 2984 2985 2986 2987 2988 2989

            # x is a Tensor variable with shape [2, 2, 2] and elements as below:
            #      [[[1.0, 2.0], [3.0, 4.0]],
            #      [[5.0, 6.0], [7.0, 8.0]]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_max(x, dim=[1, 2]) # [4.0, 8.0]
            fluid.layers.reduce_max(x, dim=[0, 1]) # [7.0, 8.0]
2990 2991 2992
    """
    helper = LayerHelper('reduce_max', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2993 2994
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
2995 2996 2997 2998 2999
    helper.append_op(
        type='reduce_max',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
3000
            'dim': dim if dim != None else [0],
3001 3002 3003 3004 3005 3006
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out


C
caoying03 已提交
3007
def reduce_min(input, dim=None, keep_dim=False, name=None):
3008
    """
Y
yangyaming 已提交
3009
    Computes the minimum of tensor elements over the given dimension.
3010 3011 3012

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
3013
        dim (list|int|None): The dimensions along which the minimum is computed.
Y
yangyaming 已提交
3014 3015 3016
            If :attr:`None`, compute the minimum over all elements of
            :attr:`input` and return a Tensor variable with a single element,
            otherwise must be in the range :math:`[-rank(input), rank(input))`.
W
whs 已提交
3017
            If :math:`dim[i] < 0`, the dimension to reduce is :math:`rank + dim[i]`.
Y
yangyaming 已提交
3018 3019
        keep_dim (bool): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
3020
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
3021 3022
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
3023 3024 3025

    Returns:
        Variable: The reduced Tensor variable.
Y
yangyaming 已提交
3026

3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037
    Examples:
        .. code-block:: python

            # x is a Tensor variable with following elements:
            #    [[0.2, 0.3, 0.5, 0.9]
            #     [0.1, 0.2, 0.6, 0.7]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_min(x)  # [0.1]
            fluid.layers.reduce_min(x, dim=0)  # [0.1, 0.2, 0.5, 0.7]
            fluid.layers.reduce_min(x, dim=-1)  # [0.2, 0.1]
            fluid.layers.reduce_min(x, dim=1, keep_dim=True)  # [[0.2], [0.1]]
W
whs 已提交
3038 3039 3040 3041 3042 3043 3044

            # x is a Tensor variable with shape [2, 2, 2] and elements as below:
            #      [[[1.0, 2.0], [3.0, 4.0]],
            #      [[5.0, 6.0], [7.0, 8.0]]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_min(x, dim=[1, 2]) # [1.0, 5.0]
            fluid.layers.reduce_min(x, dim=[0, 1]) # [1.0, 2.0]
3045 3046 3047
    """
    helper = LayerHelper('reduce_min', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
3048 3049
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
3050 3051 3052 3053 3054
    helper.append_op(
        type='reduce_min',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
3055
            'dim': dim if dim != None else [0],
3056 3057 3058 3059
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
G
guosheng 已提交
3060 3061


3062 3063 3064 3065 3066 3067
def reduce_prod(input, dim=None, keep_dim=False, name=None):
    """
    Computes the product of tensor elements over the given dimension.

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
3068
        dim (list|int|None): The dimensions along which the product is performed. If
3069 3070
            :attr:`None`, multipy all elements of :attr:`input` and return a
            Tensor variable with a single element, otherwise must be in the
W
whs 已提交
3071 3072
            range :math:`[-rank(input), rank(input))`. If :math:`dim[i] < 0`,
            the dimension to reduce is :math:`rank + dim[i]`.
3073 3074 3075
        keep_dim (bool|False): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
            than the :attr:`input` unless :attr:`keep_dim` is true.
Y
yangyaming 已提交
3076
        name(str|None): A name for this layer(optional). If set None, the
Z
zhouhanqing 已提交
3077
            layer will be named automatically.
3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091

    Returns:
        Variable: The reduced Tensor variable.

    Examples:
        .. code-block:: python

            # x is a Tensor variable with following elements:
            #    [[0.2, 0.3, 0.5, 0.9]
            #     [0.1, 0.2, 0.6, 0.7]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_prod(x)  # [0.0002268]
            fluid.layers.reduce_prod(x, dim=0)  # [0.02, 0.06, 0.3, 0.63]
            fluid.layers.reduce_prod(x, dim=-1)  # [0.027, 0.0084]
Y
yangyaming 已提交
3092
            fluid.layers.reduce_prod(x, dim=1,
Z
zhouhanqing 已提交
3093
                                     keep_dim=True)  # [[0.027], [0.0084]]
W
whs 已提交
3094 3095 3096 3097 3098 3099 3100

            # x is a Tensor variable with shape [2, 2, 2] and elements as below:
            #      [[[1.0, 2.0], [3.0, 4.0]],
            #      [[5.0, 6.0], [7.0, 8.0]]]
            # Each example is followed by the correspending output tensor.
            fluid.layers.reduce_prod(x, dim=[1, 2]) # [24.0, 1680.0]
            fluid.layers.reduce_prod(x, dim=[0, 1]) # [105.0, 384.0]
3101 3102 3103
    """
    helper = LayerHelper('reduce_prod', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
3104 3105
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
3106 3107 3108 3109 3110
    helper.append_op(
        type='reduce_prod',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
3111
            'dim': dim if dim != None else [0],
3112 3113 3114 3115 3116 3117
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out


C
caoying03 已提交
3118
def split(input, num_or_sections, dim=-1, name=None):
G
guosheng 已提交
3119
    """
C
caoying03 已提交
3120
    Split the input tensor into multiple sub-tensors.
G
guosheng 已提交
3121 3122 3123

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
C
caoying03 已提交
3124 3125 3126 3127 3128
        num_or_sections (int|list): If :attr:`num_or_sections` is an integer,
            then the integer indicates the number of equal sized sub-tensors
            that the tensor will be divided into. If :attr:`num_or_sections`
            is a list of integers, the length of list indicates the number of
            sub-tensors and the integers indicate the sizes of sub-tensors'
G
guosheng 已提交
3129
            :attr:`dim` dimension orderly.
C
caoying03 已提交
3130
        dim (int): The dimension along which to split. If :math:`dim < 0`, the
G
guosheng 已提交
3131
            dimension to split along is :math:`rank(input) + dim`.
C
caoying03 已提交
3132 3133
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
G
guosheng 已提交
3134 3135

    Returns:
D
dzhwinter 已提交
3136
        list(Variable): The list of segmented tensor variables.
G
guosheng 已提交
3137 3138 3139 3140 3141 3142 3143 3144 3145

    Examples:
        .. code-block:: python

            # x is a Tensor variable with shape [3, 9, 5]:
            x0, x1, x2 = fluid.layers.split(x, num_or_sections=3, dim=1)
            x0.shape  # [3, 3, 5]
            x1.shape  # [3, 3, 5]
            x2.shape  # [3, 3, 5]
F
stash  
fengjiayi 已提交
3146 3147
            x0, x1, x2 = fluid.layers.split(
                x, num_or_sections=[2, 3, 4], dim=1)
G
guosheng 已提交
3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176
            x0.shape  # [3, 2, 5]
            x1.shape  # [3, 3, 5]
            x2.shape  # [3, 4, 5]
    """
    helper = LayerHelper('split', **locals())
    input_shape = input.shape
    dim = (len(input_shape) + dim) if dim < 0 else dim
    if isinstance(num_or_sections, int):
        assert num_or_sections > 1, 'num_or_sections must be more than 1.'
        num = num_or_sections
    else:
        assert len(num_or_sections) < input_shape[
            dim], 'len(num_or_sections) must not be more than input.shape[dim].'
        num = len(num_or_sections)
    outs = [
        helper.create_tmp_variable(dtype=helper.input_dtype())
        for i in range(num)
    ]
    helper.append_op(
        type='split',
        inputs={'X': input},
        outputs={'Out': outs},
        attrs={
            'num': num_or_sections if isinstance(num_or_sections, int) else 0,
            'sections': num_or_sections
            if isinstance(num_or_sections, list) else [],
            'axis': dim
        })
    return outs
C
caoying03 已提交
3177 3178 3179 3180 3181 3182 3183 3184 3185


def l2_normalize(x, axis, epsilon=1e-12, name=None):
    """
    **L2 normalize Layer**

    The l2 normalize layer normalizes `x` along dimension `axis` using an L2
    norm. For a 1-D tensor (`dim` is fixed to 0), this layer computes

3186
    .. math::
3187 3188

        y = \\frac{x}{ \sqrt{\sum {x^2} + epsion }}
C
caoying03 已提交
3189 3190 3191 3192 3193

    For `x` with more dimensions, this layer independently normalizes each 1-D
    slice along dimension `axis`.

    Args:
3194
        x(Variable|list): The input tensor to l2_normalize layer.
3195
        axis(int): The axis on which to apply normalization. If `axis < 0`, \
3196 3197
            the dimension to normalization is rank(X) + axis. -1 is the
            last dimension.
3198
        epsilon(float): The epsilon value is used to avoid division by zero, \
3199
            the defalut value is 1e-10.
3200
        name(str|None): A name for this layer(optional). If set None, the layer \
3201
            will be named automatically.
C
caoying03 已提交
3202 3203

    Returns:
3204
        Variable: The output tensor variable is the same shape with `x`.
C
caoying03 已提交
3205 3206

    Examples:
3207

C
caoying03 已提交
3208 3209
        .. code-block:: python

3210 3211 3212 3213
            data = fluid.layers.data(name="data",
                                     shape=(3, 17, 13),
                                     dtype="float32")
            normed = fluid.layers.l2_normalize(x=data, axis=1)
C
caoying03 已提交
3214 3215
    """

F
fengjiayi 已提交
3216 3217
    if len(x.shape) == 1:
        axis = 0
C
caoying03 已提交
3218 3219
    helper = LayerHelper("l2_normalize", **locals())

3220 3221
    out = helper.create_tmp_variable(dtype=x.dtype)
    norm = helper.create_tmp_variable(dtype=x.dtype)
C
caoying03 已提交
3222
    helper.append_op(
3223 3224 3225 3226
        type="norm",
        inputs={"X": x},
        outputs={"Out": out,
                 "Norm": norm},
C
caoying03 已提交
3227
        attrs={
3228 3229
            "axis": 1 if axis is None else axis,
            "epsilon": epsilon,
C
caoying03 已提交
3230 3231
        })
    return out
3232 3233


3234
def matmul(x, y, transpose_x=False, transpose_y=False, name=None):
G
guosheng 已提交
3235
    """
Y
ying 已提交
3236 3237 3238 3239
    Applies matrix multiplication to two tensors.

    Currently, the input tensors' rank can be any, but when the rank of any
    inputs is bigger than 3, this two inputs' rank should be equal.
G
guosheng 已提交
3240

C
chengduoZH 已提交
3241
    The actual behavior depends on the shapes of :math:`x`, :math:`y` and the
3242
    flag values of :attr:`transpose_x`, :attr:`transpose_y`. Specifically:
G
guosheng 已提交
3243

3244 3245 3246 3247 3248
    - If a transpose flag is specified, the last two dimensions of the tensor
      are transposed. If the tensor is rank-1 of shape :math:`[D]`, then for
      :math:`x` it is treated as :math:`[1, D]` in nontransposed form and as
      :math:`[D, 1]` in transposed form, whereas for :math:`y` it is the
      opposite: It is treated as :math:`[D, 1]` in nontransposed form and as
3249
      :math:`[1, D]` in transposed form.
G
guosheng 已提交
3250

C
chengduoZH 已提交
3251
    - After transpose, the two tensors are 2-D or n-D and matrix multiplication
3252
      performs in the following way.
G
guosheng 已提交
3253

3254
      - If both are 2-D, they are multiplied like conventional matrices.
C
chengduoZH 已提交
3255
      - If either is n-D, it is treated as a stack of matrices residing in the
Y
ying 已提交
3256
        last two dimensions and a batched matrix multiply supporting broadcast
3257
        applies on the two tensors.
G
guosheng 已提交
3258

Y
ying 已提交
3259 3260
    Also note that if the raw tensor :math:`x` or :math:`y` is rank-1 and
    nontransposed, the prepended or appended dimension :math:`1` will be
C
chengduoZH 已提交
3261
    removed after matrix multiplication.
G
guosheng 已提交
3262 3263 3264

    Args:
        x (Variable): The input variable which is a Tensor or LoDTensor.
3265 3266 3267
        y (Variable): The input variable which is a Tensor or LoDTensor.
        transpose_x (bool): Whether to transpose :math:`x` before multiplication.
        transpose_y (bool): Whether to transpose :math:`y` before multiplication.
3268
        name(str|None): A name for this layer(optional). If set None, the layer
3269
            will be named automatically.
G
guosheng 已提交
3270 3271

    Returns:
3272
        Variable: The product Tensor variable.
G
guosheng 已提交
3273

G
guosheng 已提交
3274 3275 3276
    Examples:
        .. code-block:: python

3277
            # Examples to clarify shapes of the inputs and output
C
chengduoZH 已提交
3278 3279
            # x: [B, ..., M, K], y: [B, ..., K, N]
            fluid.layers.matmul(x, y)  # out: [B, ..., M, N]
Y
ying 已提交
3280

3281 3282
            # x: [B, M, K], y: [B, K, N]
            fluid.layers.matmul(x, y)  # out: [B, M, N]
Y
ying 已提交
3283

3284 3285
            # x: [B, M, K], y: [K, N]
            fluid.layers.matmul(x, y)  # out: [B, M, N]
Y
ying 已提交
3286

3287 3288
            # x: [M, K], y: [K, N]
            fluid.layers.matmul(x, y)  # out: [M, N]
Y
ying 已提交
3289 3290 3291 3292

            # x: [B, M, K], y: [K]
            fluid.layers.matmul(x, y)  # out: [B, M]

3293 3294
            # x: [K], y: [K]
            fluid.layers.matmul(x, y)  # out: [1]
3295

Y
ying 已提交
3296
            # x: [M], y: [N]
3297
            fluid.layers.matmul(x, y, True, True)  # out: [M, N]
G
guosheng 已提交
3298
    """
Y
ying 已提交
3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310

    def __check_input(x, y):
        if len(y.shape) > len(x.shape):
            raise ValueError(
                "Invalid inputs for matmul. "
                "x's rank should be always greater than or equal to y'rank.")

        x_shape = list(x.shape)
        y_shape = list(y.shape)
        if len(x_shape) == 1:
            x_shape = [1] + x_shape
        if len(y_shape) == 1:
Y
ying 已提交
3311
            y_shape = y_shape + [1]
Y
ying 已提交
3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327

        # check the inner 2 dimensions
        if transpose_x:
            x_shape[-2], x_shape[-1] = x_shape[-1], x_shape[-2]
        if transpose_y:
            y_shape[-2], y_shape[-1] = y_shape[-1], y_shape[-2]
        if x_shape[-1] != y_shape[-2]:
            raise ValueError("Invalid inputs for matmul.")

        if len(y_shape) > 2:
            for i, dim_x in enumerate(x_shape[:-2]):
                if dim_x != y_shape[i]:
                    raise ValueError("Invalid inputs for matmul.")

    __check_input(x, y)

3328
    helper = LayerHelper('matmul', **locals())
Y
ying 已提交
3329
    out = helper.create_tmp_variable(dtype=x.dtype)
G
guosheng 已提交
3330
    helper.append_op(
3331 3332 3333 3334 3335 3336 3337
        type='matmul',
        inputs={'X': x,
                'Y': y},
        outputs={'Out': out},
        attrs={'transpose_X': transpose_x,
               'transpose_Y': transpose_y})
    return out
3338 3339


3340
def topk(input, k, name=None):
Q
qingqing01 已提交
3341 3342 3343 3344
    """
    This operator is used to find values and indices of the k largest entries
    for the last dimension.

F
fengjiayi 已提交
3345
    If the input is a vector (1-D Tensor), finds the k largest entries in the vector
Q
qingqing01 已提交
3346 3347 3348 3349 3350 3351
    and outputs their values and indices as vectors. Thus values[j] is the j-th
    largest entry in input, and its index is indices[j].

    If the input is a Tensor with higher rank, this operator computes the top k
    entries along the last dimension.

F
fengjiayi 已提交
3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372
    For example:

    .. code-block:: text

        If:
            input = [[5, 4, 2, 3],
                     [9, 7, 10, 25],
                     [6, 2, 10, 1]]
            k = 2

        Then:
            The first output:
            values = [[5, 4],
                      [10, 25],
                      [6, 10]]

            The second output:
            indices = [[0, 1],
                       [2, 3],
                       [0, 2]]

Q
qingqing01 已提交
3373 3374 3375
    Args:
        input(Variable): The input variable which can be a vector or Tensor with
            higher rank.
F
fengjiayi 已提交
3376 3377
        k(int):  The number of top elements to look for along the last dimension 
                 of input.
3378
        name(str|None): A name for this layer(optional). If set None, the layer
F
fengjiayi 已提交
3379 3380
                       will be named automatically. 
                       Default: None
Q
qingqing01 已提交
3381 3382

    Returns:
F
fengjiayi 已提交
3383 3384 3385 3386
        Tuple[Variable]: A tuple with two elements. Each element is a Variable. 
        The first one is k largest elements along each last 
        dimensional slice. The second one is indices of values 
        within the last dimension of input.
Q
qingqing01 已提交
3387

F
fengjiayi 已提交
3388 3389
    Raises:
        ValueError: If k < 1 or k is not less than the last dimension of input
Q
qingqing01 已提交
3390 3391 3392 3393 3394 3395 3396

    Examples:
        .. code-block:: python

            top5_values, top5_indices = layers.topk(input, k=5)
    """
    shape = input.shape
F
fengjiayi 已提交
3397
    if k < 1 or k >= shape[-1]:
Q
qingqing01 已提交
3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414
        raise ValueError("k must be greater than 0 and less than %d." %
                         (shape[-1]))

    helper = LayerHelper("top_k", **locals())
    values = helper.create_tmp_variable(dtype=input.dtype)
    indices = helper.create_tmp_variable(dtype="int64")
    helper.append_op(
        type="top_k",
        inputs={"X": [input]},
        outputs={"Out": [values],
                 "Indices": [indices]},
        attrs={"k": k})
    values.stop_gradient = True
    indices.stop_gradient = True
    return values, indices


3415
def edit_distance(input, label, normalized=True, ignored_tokens=None):
3416
    """
Y
ying 已提交
3417 3418 3419 3420 3421 3422 3423 3424 3425
    EditDistance operator computes the edit distances between a batch of
    hypothesis strings and their references. Edit distance, also called
    Levenshtein distance, measures how dissimilar two strings are by counting
    the minimum number of operations to transform one string into anthor.
    Here the operations include insertion, deletion, and substitution.

    For example, given hypothesis string A = "kitten" and reference
    B = "sitting", the edit distance is 3 for A will be transformed into B
    at least after two substitutions and one insertion:
W
wanghaoshuang 已提交
3426

Y
ying 已提交
3427
    "kitten" -> "sitten" -> "sittin" -> "sitting"
W
wanghaoshuang 已提交
3428

3429
    The input is a LoDTensor consisting of all the hypothesis strings with
Y
ying 已提交
3430 3431
    the total number denoted by `batch_size`, and the separation is specified
    by the LoD information. And the `batch_size` reference strings are arranged
3432
    in order in the same way in the input LoDTensor.
W
wanghaoshuang 已提交
3433

3434
    The output contains the `batch_size` results and each stands for the edit
Y
ying 已提交
3435 3436
    distance for a pair of strings respectively. If Attr(normalized) is true,
    the edit distance will be divided by the length of reference string.
W
wanghaoshuang 已提交
3437

3438 3439 3440
    Args:
        input(Variable): The indices for hypothesis strings.
        label(Variable): The indices for reference strings.
3441
        normalized(bool, default True): Indicated whether to normalize the edit distance by
Y
ying 已提交
3442
                          the length of reference string.
3443
        ignored_tokens(list<int>, default None): Tokens that should be removed before
Y
ying 已提交
3444
                                     calculating edit distance.
3445
        name (str): The name of this layer. It is optional.
3446

W
wanghaoshuang 已提交
3447
    Returns:
W
wanghaoshuang 已提交
3448
        Variable: sequence-to-sequence edit distance in shape [batch_size, 1].
W
wanghaoshuang 已提交
3449 3450 3451 3452 3453

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[8], dtype='float32')
3454
            y = fluid.layers.data(name='y', shape=[7], dtype='float32')
3455
            cost = fluid.layers.edit_distance(input=x,label=y)
3456
    """
3457
    helper = LayerHelper("edit_distance", **locals())
3458

3459
    # remove some tokens from input and labels
W
wanghaoshuang 已提交
3460
    if ignored_tokens is not None and len(ignored_tokens) > 0:
3461 3462 3463 3464 3465 3466 3467
        erased_input = helper.create_tmp_variable(dtype="int64")
        erased_label = helper.create_tmp_variable(dtype="int64")

        helper.append_op(
            type="sequence_erase",
            inputs={"X": [input]},
            outputs={"Out": [erased_input]},
W
wanghaoshuang 已提交
3468
            attrs={"tokens": ignored_tokens})
3469 3470 3471 3472 3473
        input = erased_input

        helper.append_op(
            type="sequence_erase",
            inputs={"X": [label]},
W
whs 已提交
3474
            outputs={"Out": [erased_label]},
W
wanghaoshuang 已提交
3475
            attrs={"tokens": ignored_tokens})
3476 3477
        label = erased_label

3478 3479
    # edit distance op
    edit_distance_out = helper.create_tmp_variable(dtype="int64")
3480
    sequence_num = helper.create_tmp_variable(dtype="int64")
3481 3482 3483 3484
    helper.append_op(
        type="edit_distance",
        inputs={"Hyps": [input],
                "Refs": [label]},
3485 3486
        outputs={"Out": [edit_distance_out],
                 "SequenceNum": [sequence_num]},
3487 3488
        attrs={"normalized": normalized})

3489
    return edit_distance_out, sequence_num
3490 3491 3492 3493 3494


def ctc_greedy_decoder(input, blank, name=None):
    """
    This op is used to decode sequences by greedy policy by below steps:
Y
yi.wu 已提交
3495

Y
ying 已提交
3496 3497 3498 3499
    1. Get the indexes of max value for each row in input. a.k.a.
       numpy.argmax(input, axis=0).
    2. For each sequence in result of step1, merge repeated tokens between two
       blanks and delete all blanks.
3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516

    A simple example as below:

    .. code-block:: text

        Given:

        input.data = [[0.6, 0.1, 0.3, 0.1],
                      [0.3, 0.2, 0.4, 0.1],
                      [0.1, 0.5, 0.1, 0.3],
                      [0.5, 0.1, 0.3, 0.1],

                      [0.5, 0.1, 0.3, 0.1],
                      [0.2, 0.2, 0.2, 0.4],
                      [0.2, 0.2, 0.1, 0.5],
                      [0.5, 0.1, 0.3, 0.1]]

3517
        input.lod = [[4, 4]]
3518 3519 3520 3521 3522 3523 3524

        Then:

        output.data = [[2],
                       [1],
                       [3]]

3525
        output.lod = [[2, 1]]
3526 3527 3528

    Args:

Y
ying 已提交
3529 3530 3531 3532 3533 3534 3535 3536 3537
        input(Variable): (LoDTensor<float>), the probabilities of
                         variable-length sequences, which is a 2-D Tensor with
                         LoD information. It's shape is [Lp, num_classes + 1],
                         where Lp is the sum of all input sequences' length and
                         num_classes is the true number of classes. (not
                         including the blank label).
        blank(int): the blank label index of Connectionist Temporal
                    Classification (CTC) loss, which is in thehalf-opened
                    interval [0, num_classes + 1).
3538
        name (str): The name of this layer. It is optional.
3539 3540

    Returns:
3541
        Variable: CTC greedy decode result. If all the sequences in result were
3542
        empty, the result LoDTensor will be [-1] with LoD [[]] and dims [1, 1].
3543 3544 3545 3546 3547

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[8], dtype='float32')
W
wanghaoshuang 已提交
3548

3549
            cost = fluid.layers.ctc_greedy_decoder(input=x, blank=0)
W
wanghaoshuang 已提交
3550
    """
3551
    helper = LayerHelper("ctc_greedy_decoder", **locals())
Q
qingqing01 已提交
3552
    _, topk_indices = topk(input, k=1)
3553 3554 3555 3556 3557 3558

    # ctc align op
    ctc_out = helper.create_tmp_variable(dtype="int64")
    helper.append_op(
        type="ctc_align",
        inputs={"Input": [topk_indices]},
W
wanghaoshuang 已提交
3559
        outputs={"Output": [ctc_out]},
3560 3561
        attrs={"merge_repeated": True,
               "blank": blank})
3562
    return ctc_out
3563 3564


F
fengjiayi 已提交
3565
def warpctc(input, label, blank=0, norm_by_times=False):
W
wanghaoshuang 已提交
3566
    """
3567 3568
    An operator integrating the open source Warp-CTC library
    (https://github.com/baidu-research/warp-ctc)
W
wanghaoshuang 已提交
3569
    to compute Connectionist Temporal Classification (CTC) loss.
3570 3571
    It can be aliased as softmax with CTC, since a native softmax activation is
    interated to the Warp-CTC library, to to normlize values for each row of the
W
wanghaoshuang 已提交
3572 3573 3574
    input tensor.

    Args:
3575
       input (Variable): The unscaled probabilities of variable-length sequences,
W
wanghaoshuang 已提交
3576 3577 3578 3579
         which is a 2-D Tensor with LoD information.
         It's shape is [Lp, num_classes + 1], where Lp is the sum of all input
         sequences' length and num_classes is the true number of classes.
         (not including the blank label).
3580 3581 3582 3583
       label (Variable): The ground truth of variable-length sequence, 
         which is a 2-D Tensor with LoD information. It is of the shape [Lg, 1],
         where Lg is th sum of all labels' length.
       blank (int, default 0): The blank label index of Connectionist
W
wanghaoshuang 已提交
3584 3585
         Temporal Classification (CTC) loss, which is in the
         half-opened interval [0, num_classes + 1).
3586 3587 3588 3589
       norm_by_times(bool, default false): Whether to normalize the gradients 
         by the number of time-step, which is also the sequence's length. 
         There is no need to normalize the gradients if warpctc layer was 
         follewed by a mean_op.
W
wanghaoshuang 已提交
3590 3591

    Returns:
3592 3593
        Variable: The Connectionist Temporal Classification (CTC) loss,
        which is a 2-D Tensor of the shape [batch_size, 1].
W
wanghaoshuang 已提交
3594 3595

    Examples:
3596

W
wanghaoshuang 已提交
3597
        .. code-block:: python
3598

3599 3600 3601
            label = fluid.layers.data(shape=[11, 8], dtype='float32', lod_level=1)
            predict = fluid.layers.data(shape=[11, 1], dtype='float32')
            cost = fluid.layers.warpctc(input=predict, label=label)
W
wanghaoshuang 已提交
3602 3603

    """
F
fengjiayi 已提交
3604
    helper = LayerHelper('warpctc', **locals())
W
wanghaoshuang 已提交
3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615
    loss_out = helper.create_tmp_variable(dtype=input.dtype)
    grad_out = helper.create_tmp_variable(dtype=input.dtype)
    helper.append_op(
        type='warpctc',
        inputs={'Logits': [input],
                'Label': [label]},
        outputs={'WarpCTCGrad': [grad_out],
                 'Loss': [loss_out]},
        attrs={'blank': blank,
               'norm_by_times': norm_by_times})
    return loss_out
3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630


def sequence_reshape(input, new_dim):
    """
    **Sequence Reshape Layer**

    This layer will rearrange the input sequences. The new dimension is set by
    user. Length of each sequence is computed according to original length,
    original dimension and new dimension. The following example will help to
    illustrate the function of this layer:

    .. code-block:: text

        x is a LoDTensor:
            x.lod  = [[0, 2, 6]]
3631 3632 3633
            x.data = [[1,  2], [3,  4],
                      [5,  6], [7,  8],
                      [9, 10], [11, 12]]
3634 3635 3636 3637 3638
            x.dims = [6, 2]

        set new_dim = 4

        then out is a LoDTensor:
3639

3640
            out.lod  = [[0, 1, 3]]
3641 3642 3643 3644

            out.data = [[1,  2,  3,  4],
                        [5,  6,  7,  8],
                        [9, 10, 11, 12]]
3645 3646 3647 3648 3649 3650 3651
            out.dims = [3, 4]

    Currently, only 1-level LoDTensor is supported and please make sure
    (original length * original dimension) can be divided by new dimension with
    no remainder for each sequence.

    Args:
3652 3653 3654

       input (Variable): A 2-D LoDTensor with shape being [N, M] where M for dimension.
       new_dim (int): New dimension that the input LoDTensor is reshaped to.
3655 3656

    Returns:
3657

3658 3659 3660 3661 3662
        Variable: Reshaped LoDTensor according to new dimension.

    Examples:
        .. code-block:: python

3663
            x = fluid.layers.data(shape=[5, 20], dtype='float32', lod_level=1)
3664
            x_reshaped = fluid.layers.sequence_reshape(input=x, new_dim=10)
3665 3666 3667 3668 3669 3670 3671 3672 3673
    """
    helper = LayerHelper('sequence_reshape', **locals())
    out = helper.create_tmp_variable(helper.input_dtype())
    helper.append_op(
        type='sequence_reshape',
        inputs={'X': [input]},
        outputs={'Out': [out]},
        attrs={'new_dim': new_dim})
    return out
Y
ying 已提交
3674 3675


3676 3677 3678 3679
# FIXME(wuyi): let docstring_checker.py understand @autodoc.
# For now, the comments in c++ use types like Tensor, but in python side
# the type is often "Variable", and arguments may vary.
@templatedoc(op_type="nce")
Y
Yang Yu 已提交
3680 3681 3682 3683 3684 3685 3686
def nce(input,
        label,
        num_total_classes,
        sample_weight=None,
        param_attr=None,
        bias_attr=None,
        num_neg_samples=None):
3687 3688 3689 3690 3691 3692 3693
    """
    ${comment}

    Args:
        input (Variable): input variable.
        label (Variable): label.
        num_total_classes (int):${num_total_classes_comment}
3694 3695 3696
        sample_weight (Variable|None): A Variable of shape [batch_size, 1] 
            storing a weight for each sample. The default weight for each 
            sample is 1.0.
3697 3698 3699
        param_attr (ParamAttr|None): attributes for parameter
        bias_attr (ParamAttr|None): attributes for bias
        num_neg_samples (int): ${num_neg_samples_comment}
F
fengjiayi 已提交
3700

3701
    Returns:
Y
Yibing Liu 已提交
3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728
        Variable: The output nce loss.

    Examples:
        .. code-block:: python

            window_size = 5
            words = []
            for i in xrange(window_size):
                words.append(layers.data(
                    name='word_{0}'.format(i), shape=[1], dtype='int64'))

            dict_size = 10000
            label_word = int(window_size / 2) + 1

            embs = []
            for i in xrange(window_size):
                if i == label_word:
                    continue

                emb = layers.embedding(input=words[i], size=[dict_size, 32],
                                       param_attr='emb.w', is_sparse=True)
                embs.append(emb)

            embs = layers.concat(input=embs, axis=1)
            loss = layers.nce(input=embs, label=words[label_word],
                          num_total_classes=dict_size, param_attr='nce.w',
                          bias_attr='nce.b')
3729
    """
Y
Yang Yu 已提交
3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748
    helper = LayerHelper('nce', **locals())
    assert isinstance(input, Variable)
    dim = input.shape[1]
    assert isinstance(label, Variable)
    num_true_class = label.shape[1]
    w = helper.create_parameter(
        attr=helper.param_attr,
        shape=[num_total_classes, dim],
        is_bias=False,
        dtype=input.dtype)
    b = helper.create_parameter(
        attr=helper.bias_attr,
        shape=[num_total_classes, 1],
        is_bias=True,
        dtype=input.dtype)
    cost = helper.create_tmp_variable(dtype=input.dtype)
    sample_logits = helper.create_tmp_variable(dtype=input.dtype)
    sample_labels = helper.create_tmp_variable(dtype=label.dtype)

Y
Yang Yu 已提交
3749 3750 3751 3752 3753 3754 3755 3756 3757
    if num_neg_samples is None:
        num_neg_samples = 10
    else:
        num_neg_samples = int(num_neg_samples)

    attrs = {
        'num_total_classes': int(num_total_classes),
        'num_neg_samples': num_neg_samples
    }
Y
Yang Yu 已提交
3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773

    helper.append_op(
        type='nce',
        inputs={
            'Input': input,
            'Label': label,
            'Weight': w,
            'Bias': b,
            'SampleWeight': sample_weight if sample_weight is not None else []
        },
        outputs={
            'Cost': cost,
            'SampleLogits': sample_logits,
            'SampleLabels': sample_labels
        },
        attrs=attrs)
Y
Yang Yu 已提交
3774
    return cost / (num_neg_samples + 1)
3775 3776


Y
fix ci.  
ying 已提交
3777
def transpose(x, perm, name=None):
Y
ying 已提交
3778 3779 3780 3781 3782 3783 3784
    """
    Permute the dimensions of `input` according to `perm`.

    The `i`-th dimension  of the returned tensor will correspond to the
    perm[i]-th dimension of `input`.

    Args:
3785 3786 3787
        x (Variable): The input Tensor.
        perm (list): A permutation of the dimensions of `input`.
        name (str): The name of this layer. It is optional.
Y
ying 已提交
3788 3789 3790 3791 3792 3793 3794 3795

    Returns:
        Variable: A transposed Tensor.

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[5, 10, 15], dtype='float32')
Y
fix ci.  
ying 已提交
3796
            x_transposed = layers.transpose(x, perm=[1, 0, 2])
Y
ying 已提交
3797 3798
    """

Y
fix ci.  
ying 已提交
3799
    if len(perm) != len(x.shape):
Y
ying 已提交
3800 3801 3802
        raise ValueError(
            "Input(perm) is the permutation of dimensions of Input(input). "
            "It's length shoud be equal to Input(input)'s rank.")
Y
ying 已提交
3803 3804 3805 3806 3807 3808
    for idx, dim in enumerate(perm):
        if dim >= len(x.shape):
            raise ValueError(
                "Each element in perm should be less than x's rank. "
                "%d-th element in perm is %d which accesses x's rank %d." %
                (idx, perm[idx], len(x.shape)))
Y
ying 已提交
3809 3810

    helper = LayerHelper('transpose', **locals())
Y
fix ci.  
ying 已提交
3811
    out = helper.create_tmp_variable(x.dtype)
Y
ying 已提交
3812 3813
    helper.append_op(
        type='transpose',
Y
fix ci.  
ying 已提交
3814
        inputs={'X': [x]},
Y
ying 已提交
3815 3816 3817
        outputs={'Out': [out]},
        attrs={'axis': perm})
    return out
3818 3819


3820
def im2sequence(input, filter_size=1, stride=1, padding=0, name=None):
3821
    """
3822 3823 3824 3825 3826 3827 3828
    Extracts image patches from the input tensor to form a tensor of shape
    {input.batch_size * output_height * output_width, filter_size_H *
    filter_size_W * input.channels} which is similar with im2col.
    This op use filter / kernel to scan images and convert these images to
    sequences. After expanding, the number of time step are
    output_height * output_width for an image, in which output_height and
    output_width are calculated by below equation:
3829 3830 3831 3832 3833 3834 3835 3836 3837 3838

    .. math::

        output\_size = 1 + \
            (2 * padding + img\_size - block\_size + stride - 1) / stride

    And the dimension of each time step is block_y * block_x * input.channels.

    Args:
        input (Variable): The input should be a tensor in NCHW format.
W
wanghaoshuang 已提交
3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856

        filter_size(int|tuple|None): The filter size. If filter_size is a tuple,
            it must contain two integers, (filter_size_H, filter_size_W).
            Otherwise, the filter will be a square.

        stride(int|tuple): The stride size. If stride is a tuple, it must
            contain two integers, (stride_H, stride_W). Otherwise, the
            stride_H = stride_W = stride. Default: stride = 1.

        padding(int|tuple): The padding size. If padding is a tuple, it can
            contain two integers like (padding_H, padding_W) which means
            padding_up = padding_down = padding_H and
            padding_left = padding_right = padding_W. Or it can use
            (padding_up, padding_left, padding_down, padding_right) to indicate
            paddings of four direction. Otherwise, a scalar padding means
            padding_up = padding_down = padding_left = padding_right = padding
            Default: padding = 0.

3857 3858 3859
        name (int): The name of this layer. It is optional.

    Returns:
W
wanghaoshuang 已提交
3860 3861 3862 3863 3864
        output: The output is a LoDTensor with shape
        {input.batch_size * output_height * output_width,
        filter_size_H * filter_size_W * input.channels}.
        If we regard output as a matrix, each row of this matrix is
        a step of a sequence.
3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891

    Examples:

        .. code-block:: text

            Given:

            x = [[[[ 6.  2.  1.]
                   [ 8.  3.  5.]
                   [ 0.  2.  6.]]

                  [[ 2.  4.  4.]
                   [ 6.  3.  0.]
                   [ 6.  4.  7.]]]

                 [[[ 6.  7.  1.]
                   [ 5.  7.  9.]
                   [ 2.  4.  8.]]

                  [[ 1.  2.  1.]
                   [ 1.  3.  5.]
                   [ 9.  0.  8.]]]]

            x.dims = {2, 2, 3, 3}

            And:

W
wanghaoshuang 已提交
3892 3893 3894
            filter = [2, 2]
            stride = [1, 1]
            padding = [0, 0]
3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908

            Then:

            output.data = [[ 6.  2.  8.  3.  2.  4.  6.  3.]
                           [ 2.  1.  3.  5.  4.  4.  3.  0.]
                           [ 8.  3.  0.  2.  6.  3.  6.  4.]
                           [ 3.  5.  2.  6.  3.  0.  4.  7.]
                           [ 6.  7.  5.  7.  1.  2.  1.  3.]
                           [ 7.  1.  7.  9.  2.  1.  3.  5.]
                           [ 5.  7.  2.  4.  1.  3.  9.  0.]
                           [ 7.  9.  4.  8.  3.  5.  0.  8.]]

            output.dims = {8, 9}

3909
            output.lod = [[4, 4]]
3910

D
dzhwinter 已提交
3911
     Examples:
3912 3913 3914

        .. code-block:: python

3915 3916
            output = fluid.layers.im2sequence(
                input=layer, stride=[1, 1], filter_size=[2, 2])
3917 3918

    """
W
wanghaoshuang 已提交
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929

    if isinstance(filter_size, int):
        filter_size = [filter_size, filter_size]
    if isinstance(stride, int):
        stride = [stride, stride]
    if isinstance(padding, int):
        padding = [padding, padding]
    if len(padding) == 2:
        padding.append(padding[0])
        padding.append(padding[1])

3930
    helper = LayerHelper('im2sequence', **locals())
3931 3932
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
    helper.append_op(
3933
        type='im2sequence',
3934 3935 3936
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
wanghaoshuang 已提交
3937 3938 3939
            'kernels': filter_size,
            'strides': stride,
            'paddings': padding,
3940 3941
        })
    return out
3942 3943


Y
yuyang18 已提交
3944
@templatedoc()
3945
def row_conv(input, future_context_size, param_attr=None, act=None):
Y
yuyang18 已提交
3946 3947
    """
    ${comment}
3948 3949

    Args:
Y
yuyang18 已提交
3950
        input (${x_type}): ${x_comment}.
Y
yangyaming 已提交
3951 3952
        future_context_size (int): Future context size. Please note, the shape
            of convolution kernel is [future_context_size + 1, D].
3953 3954 3955 3956 3957
        param_attr (ParamAttr): Attributes of parameters, including
            name, initializer etc.
        act (str): Non-linear activation to be applied to output variable.

    Returns:
Y
yuyang18 已提交
3958
        ${out_comment}.
3959 3960

    Examples:
Y
yuyang18 已提交
3961 3962 3963 3964
        >>> import paddle.fluid as fluid
        >>> x = fluid.layers.data(name='x', shape=[16],
        >>>                        dtype='float32', lod_level=1)
        >>> out = fluid.layers.row_conv(input=x, future_context_size=2)
3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
    """
    helper = LayerHelper('row_conv', **locals())
    dtype = helper.input_dtype()
    filter_shape = [future_context_size + 1, input.shape[1]]
    filter_param = helper.create_parameter(
        attr=helper.param_attr, shape=filter_shape, dtype=dtype)
    out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type='row_conv',
        inputs={'X': [input],
                'Filter': [filter_param]},
        outputs={'Out': [out]})
Y
yangyaming 已提交
3977
    return helper.append_activation(out)
3978 3979


Y
yuyang18 已提交
3980
@templatedoc()
3981 3982
def multiplex(inputs, index):
    """
Y
yuyang18 已提交
3983 3984 3985 3986 3987 3988 3989
    ${comment}

    >>> import paddle.fluid as fluid
    >>> x1 = fluid.layers.data(name='x1', shape=[4], dtype='float32')
    >>> x2 = fluid.layers.data(name='x2', shape=[4], dtype='float32')
    >>> index = fluid.layers.data(name='index', shape=[1], dtype='int32')
    >>> out = fluid.layers.multiplex(inputs=[x1, x2], index=index)
3990 3991

    Args:
Y
yuyang18 已提交
3992 3993
       inputs (list): ${x_comment}.
       index (${ids_type}): ${ids_comment}.
3994 3995

    Returns:
Y
yuyang18 已提交
3996
        ${out_comment}.
3997 3998
    """
    helper = LayerHelper('multiplex', **locals())
Y
yangyaming 已提交
3999 4000 4001 4002 4003 4004

    if not isinstance(inputs, list) and len(inputs) < 2:
        raise ValueError("inputs should be a list object and contains at least "
                         "2 elements.")

    out = helper.create_tmp_variable(inputs[0].dtype)
4005 4006 4007 4008 4009 4010
    helper.append_op(
        type='multiplex',
        inputs={'X': inputs,
                'Ids': index},
        outputs={'Out': [out]})
    return out
4011 4012 4013 4014 4015


def softmax_with_cross_entropy(logits, label, soft_label=False):
    """
    **Softmax With Cross Entropy Operator.**
4016

4017 4018 4019 4020
    Cross entropy loss with softmax is used as the output layer extensively. This
    operator computes the softmax normalized values for each row of the input
    tensor, after which cross-entropy loss is computed. This provides a more
    numerically stable gradient.
4021

4022 4023 4024
    Because this operator performs a softmax on logits internally, it expects
    unscaled logits. This operator should not be used with the output of
    softmax operator since that would produce incorrect results.
4025

4026 4027 4028
    When the attribute soft_label is set false, this operators expects mutually
    exclusive hard labels, each sample in a batch is in exactly one class with a
    probability of 1.0. Each sample in the batch will have a single label.
4029

4030
    The equation is as follows:
4031

4032
    1) Hard label (one-hot label, so every sample has exactly one class)
4033

4034 4035 4036 4037
    .. math::

        loss_j =  -\\text{logit}_{label_j} +
        \\log\\left(\\sum_{i=0}^{K}\\exp(\\text{logit}_i)\\right), j = 1,..., K
4038

4039 4040 4041
    2) Soft label (each sample can have a distribution over all classes)

    .. math::
4042

4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063
        loss_j =  -\\sum_{i=0}^{K}\\text{label}_i
        \\left(\\text{logit}_i - \\log\\left(\\sum_{i=0}^{K}
        \\exp(\\text{logit}_i)\\right)\\right), j = 1,...,K

    Args:
        logits (Variable): The unscaled log probabilities, which is a 2-D tensor
            with shape [N x K]. N is the batch_size, and K is the class number.
        label (Variable): The ground truth which is a 2-D tensor. If soft_label
            is set to false, Label is a Tensor<int64> with shape [N x 1]. If
            soft_label is set to true, Label is a Tensor<float/double> with
        soft_label (bool): A flag to indicate whether to interpretate the given
            labels as soft labels. By default, `soft_label` is set to False.
    Returns:
        Variable: The cross entropy loss is a 2-D tensor with shape [N x 1].

    Examples:
        .. code-block:: python

            data = fluid.layers.data(name='data', shape=[128], dtype='float32')
            label = fluid.layers.data(name='label', shape=[1], dtype='int64')
            fc = fluid.layers.fc(input=data, size=100)
F
stash  
fengjiayi 已提交
4064 4065
            out = fluid.layers.softmax_with_cross_entropy(
                logits=fc, label=label)
4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
    """
    helper = LayerHelper('softmax_with_cross_entropy', **locals())
    softmax = helper.create_tmp_variable(dtype=logits.dtype)
    loss = helper.create_tmp_variable(dtype=logits.dtype)
    helper.append_op(
        type='softmax_with_cross_entropy',
        inputs={'Logits': logits,
                'Label': label},
        outputs={'Softmax': softmax,
                 'Loss': loss},
        attrs={'soft_label': soft_label})
    return loss


def smooth_l1(x, y, inside_weight=None, outside_weight=None, sigma=None):
    """
Y
Yibing Liu 已提交
4082 4083
    This layer computes the smooth L1 loss for Variable :attr:`x` and :attr:`y`.
    It takes the first dimension of :attr:`x` and :attr:`y` as batch size.
Q
qingqing01 已提交
4084
    For each instance, it computes the smooth L1 loss element by element first
4085 4086
    and then sums all the losses. So the shape of ouput Variable is 
    [batch_size, 1].
4087

4088 4089
    Args:
        x (Variable): A tensor with rank at least 2. The input value of smooth
Q
qingqing01 已提交
4090
            L1 loss op with shape [batch_size, dim1, ..., dimN].
4091
        y (Variable): A tensor with rank at least 2. The target value of smooth
Y
Yibing Liu 已提交
4092
            L1 loss op with same shape as :attr:`x`.
4093
        inside_weight (Variable|None):  A tensor with rank at least 2. This
Y
Yibing Liu 已提交
4094 4095 4096
            input is optional and should have same shape with :attr:`x`. If 
            provided, the result of (:attr:`x` - :attr:`y`) will be multiplied 
            by this tensor element by element.
4097
        outside_weight (Variable|None): A tensor with rank at least 2. This
Y
Yibing Liu 已提交
4098 4099 4100
            input is optional and should have same shape with :attr:`x`. If 
            provided, the out smooth L1 loss will be multiplied by this tensor 
            element by element.
4101 4102 4103
        sigma (float|None): Hyper parameter of smooth L1 loss layer. A float 
           scalar with default value 1.0.

4104
    Returns:
4105
        Variable: The output smooth L1 loss with shape [batch_size, 1].
4106 4107 4108 4109 4110

    Examples:
        .. code-block:: python

            data = fluid.layers.data(name='data', shape=[128], dtype='float32')
F
stash  
fengjiayi 已提交
4111 4112
            label = fluid.layers.data(
                name='label', shape=[100], dtype='float32')
4113
            fc = fluid.layers.fc(input=data, size=100)
F
fengjiayi 已提交
4114
            out = fluid.layers.smooth_l1(x=fc, y=label)
4115
    """
4116

4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131
    helper = LayerHelper('smooth_l1_loss', **locals())
    diff = helper.create_tmp_variable(dtype=x.dtype)
    loss = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type='smooth_l1_loss',
        inputs={
            'X': x,
            'Y': y,
            'InsideWeight': inside_weight,
            'OutsideWeight': outside_weight
        },
        outputs={'Diff': diff,
                 'Out': loss},
        attrs={'sigma': sigma})
    return loss
4132 4133 4134 4135


def one_hot(input, depth):
    """
Y
Yibing Liu 已提交
4136
    This layer creates the one-hot representations for input indices.
4137 4138

    Args:
Y
Yibing Liu 已提交
4139 4140
        input(Variable): Input indices, last dimension must be 1.
        depth(scalar): An interger defining the depth of the one-hot dimension.
4141 4142

    Returns:
Y
Yibing Liu 已提交
4143
        Variable: The one-hot representations of input.
4144 4145

    Examples:
C
caoying03 已提交
4146
        .. code-block:: python
F
fengjiayi 已提交
4147

Y
Yibing Liu 已提交
4148 4149
            label = layers.data(name="label", shape=[1], dtype="float32")
            one_hot_label = layers.one_hot(input=label, depth=10)
4150 4151 4152 4153 4154 4155 4156 4157 4158
    """
    helper = LayerHelper("one_hot", **locals())
    one_hot_out = helper.create_tmp_variable(dtype='float32')
    helper.append_op(
        type="one_hot",
        inputs={'X': input},
        attrs={'depth': depth},
        outputs={'Out': one_hot_out})
    return one_hot_out
Y
Yu Yang 已提交
4159 4160


Y
Yu Yang 已提交
4161
def autoincreased_step_counter(counter_name=None, begin=1, step=1):
Y
Yu Yang 已提交
4162
    """
Y
yi.wu 已提交
4163 4164 4165
    Create an auto-increase variable
    which will be automatically increased by 1 every mini-batch
    Return the run counter of the main program, default is started from 1.
Y
Yu Yang 已提交
4166 4167 4168 4169 4170 4171

    Args:
        counter_name(str): The counter name, default is '@STEP_COUNTER@'.
        begin(int): The first value of this counter.
        step(int): The increment step between each execution.

4172 4173
    Returns:
        Variable: The global run counter.
Y
yi.wu 已提交
4174 4175 4176 4177 4178 4179

    Examples:
        .. code-block:: python

           global_step = fluid.layers.autoincreased_step_counter(
               counter_name='@LR_DECAY_COUNTER@', begin=begin, step=1)
Y
Yu Yang 已提交
4180 4181
    """
    helper = LayerHelper('global_step_counter')
Y
Yu Yang 已提交
4182 4183
    if counter_name is None:
        counter_name = '@STEP_COUNTER@'
Y
Yu Yang 已提交
4184 4185 4186 4187 4188
    counter, is_new_var = helper.create_or_get_global_variable(
        name=counter_name, dtype='int64', shape=[1], persistable=True)
    if is_new_var:
        helper.set_variable_initializer(
            counter, initializer=Constant(
Y
Yu Yang 已提交
4189
                value=begin - 1, force_cpu=True))
Y
Yu Yang 已提交
4190 4191 4192
        helper.main_program.global_block().prepend_op(
            type='increment',
            inputs={'X': [counter]},
Y
Yu Yang 已提交
4193 4194
            outputs={'Out': [counter]},
            attrs={'step': float(step)})
Y
Yu Yang 已提交
4195 4196 4197
        counter.stop_gradient = True

    return counter
Y
yangyaming 已提交
4198 4199


4200
def reshape(x, shape, actual_shape=None, act=None, inplace=True, name=None):
C
caoying03 已提交
4201
    """
C
caoying03 已提交
4202 4203
    Gives a new shape to the input Tensor without changing its data.

4204 4205 4206 4207 4208
    The target shape can be given by :attr:`shape` or :attr:`actual_shape`.
    :attr:`shape` is a list of integer while :attr:`actual_shape` is a tensor
    variable. :attr:`actual_shape` has a higher priority than :attr:`shape`
    if it is provided, while :attr:`shape` still should be set correctly to
    gurantee shape inference in compile-time.
C
caoying03 已提交
4209

4210
    Some tricks exist when specifying the target shape.
C
caoying03 已提交
4211

4212 4213 4214 4215
    1. -1 means the value of this dimension is inferred from the total element
    number of x and remaining dimensions. Thus one and only one dimension can
    be set -1.

4216
    2. 0 means the actual dimension value is going to be copied from the
4217 4218 4219 4220
    corresponding dimension of x. The indice of 0s in shape can not exceed
    Rank(X).

    Here are some examples to explain it.
C
caoying03 已提交
4221 4222

    1. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
W
wanghaoshuang 已提交
4223
    is [6, 8], the reshape operator will transform x into a 2-D tensor with
4224
    shape [6, 8] and leaving x's data unchanged.
C
caoying03 已提交
4225

4226
    2. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
4227 4228
    specified is [2, 3, -1, 2], the reshape operator will transform x into a
    4-D tensor with shape [2, 3, 4, 2] and leaving x's data unchanged. In this
W
wanghaoshuang 已提交
4229 4230
    case, one dimension of the target shape is set to -1, the value of this
    dimension is inferred from the total element number of x and remaining
4231
    dimensions.
C
caoying03 已提交
4232

4233
    3. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
4234 4235 4236 4237
    is [-1, 0, 3, 2], the reshape operator will transform x into a 4-D tensor
    with shape [2, 4, 3, 2] and leaving x's data unchanged. In this case,
    besides -1, 0 means the actual dimension value is going to be copied from
    the corresponding dimension of x.
C
caoying03 已提交
4238 4239

    Args:
4240
        x(variable): The input tensor.
C
caoying03 已提交
4241 4242
        shape(list): The new shape. At most one dimension of the new shape can
                     be -1.
4243 4244 4245 4246 4247
        actual_shape(variable): An optional input. If provided, reshape
                                according to this given shape rather than
                                :attr:`shape` specifying shape. That is to
                                say :attr:`actual_shape` has a higher priority
                                than :attr:`shape`.
C
caoying03 已提交
4248 4249 4250 4251
        act (str): The non-linear activation to be applied to output variable.
        inplace(bool): If this flag is set true, a new output tensor is created
                       whose data is copied from input x, otherwise the output
                       shares data with input without copying.
4252
        name (str): The name of this layer. It is optional.
C
caoying03 已提交
4253

4254 4255
    Returns:
        Variable: The output tensor.
C
caoying03 已提交
4256 4257 4258

    Examples:
        .. code-block:: python
G
guosheng 已提交
4259

4260
            data = fluid.layers.data(
4261
                name='data', shape=[2, 4, 6], dtype='float32')
C
caoying03 已提交
4262
            reshaped = fluid.layers.reshape(
4263
                x=data, shape=[-1, 0, 3, 2], act='tanh', inplace=True)
C
caoying03 已提交
4264 4265 4266 4267 4268
    """

    if not (isinstance(shape, list) or isinstance(shape, tuple)):
        raise ValueError("Input shape must be a python lsit or tuple.")

4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283
    # Validate the shape
    unk_dim_idx = -1
    for dim_idx, dim_size in enumerate(shape):
        if dim_size == -1:
            assert unk_dim_idx == -1, (
                "Only one dimension in shape can be unknown.")
            unk_dim_idx = dim_idx
        elif dim_size == 0:
            assert dim_idx < len(x.shape), (
                "The indice of 0s in shape can not exceed Rank(X).")
        else:
            assert dim_size > 0, (
                "Each dimension size given in shape must not be negtive "
                "except one unknown dimension.")

C
caoying03 已提交
4284 4285 4286 4287
    helper = LayerHelper("reshape", **locals())
    reshaped = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type="reshape",
4288 4289 4290
        inputs={"X": x,
                "Shape": actual_shape}
        if isinstance(actual_shape, Variable) else {"X": x},
C
caoying03 已提交
4291 4292 4293 4294 4295
        attrs={"shape": shape,
               "inplace": inplace},
        outputs={"Out": reshaped})

    return helper.append_activation(reshaped)
4296 4297


Y
yangyaming 已提交
4298
def lod_reset(x, y=None, target_lod=None):
Y
yangyaming 已提交
4299
    """
Y
Yibing Liu 已提交
4300 4301 4302 4303 4304 4305
    Set LoD of :attr:`x` to a new one specified by :attr:`y` or
    :attr:`target_lod`. When :attr:`y` provided, :attr:`y.lod` would be 
    considered as target LoD first, otherwise :attr:`y.data` would be 
    considered as target LoD. If :attr:`y` is not provided, target LoD should 
    be specified by :attr:`target_lod`. If target LoD is specified by 
    :attr:`Y.data` or :attr:`target_lod`, only one level LoD is supported.
Y
yangyaming 已提交
4306 4307 4308 4309 4310 4311

    .. code-block:: text

        * Example 1:

            Given a 1-level LoDTensor x:
4312
                x.lod =  [[ 2,           3,                   1 ]]
Y
yangyaming 已提交
4313 4314 4315
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

4316
            target_lod: [4, 2]
Y
yangyaming 已提交
4317 4318

            then we get a 1-level LoDTensor:
4319
                out.lod =  [[4,                          2]]
Y
yangyaming 已提交
4320 4321 4322 4323 4324 4325
                out.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                out.dims = [6, 1]

        * Example 2:

            Given a 1-level LoDTensor x:
4326
                x.lod =  [[2,            3,                   1]]
Y
yangyaming 已提交
4327 4328 4329 4330
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

            y is a Tensor:
4331
                y.data = [[2, 4]]
Y
yangyaming 已提交
4332 4333 4334
                y.dims = [1, 3]

            then we get a 1-level LoDTensor:
4335
                out.lod =  [[2,            4]]
Y
yangyaming 已提交
4336 4337 4338 4339 4340 4341
                out.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                out.dims = [6, 1]

        * Example 3:

            Given a 1-level LoDTensor x:
4342
                x.lod =  [[2,            3,                   1]]
Y
yangyaming 已提交
4343 4344 4345 4346
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

            y is a 2-level LoDTensor:
4347
                y.lod =  [[2, 2], [2, 2, 1, 1]]
Y
yangyaming 已提交
4348 4349 4350 4351
                y.data = [[1.1], [2.1], [3.1], [4.1], [5.1], [6.1]]
                y.dims = [6, 1]

            then we get a 2-level LoDTensor:
4352
                out.lod =  [[2, 2], [2, 2, 1, 1]]
Y
yangyaming 已提交
4353 4354 4355 4356 4357
                out.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                out.dims = [6, 1]

    Args:
        x (Variable): Input variable which could be a Tensor or LodTensor.
Y
Yibing Liu 已提交
4358 4359
        y (Variable|None): If provided, output's LoD would be derived 
                           from :attr:`y`.
Y
yangyaming 已提交
4360
        target_lod (list|tuple|None): One level LoD which should be considered
Y
Yibing Liu 已提交
4361
                                      as target LoD when :attr:`y` not provided.
Y
yangyaming 已提交
4362 4363

    Returns:
Y
Yibing Liu 已提交
4364
        Variable: Output variable with LoD specified by this layer.
Y
yangyaming 已提交
4365 4366

    Raises:
Y
Yibing Liu 已提交
4367
        ValueError: If :attr:`y` and :attr:`target_lod` are both None.
Y
yangyaming 已提交
4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391

    Examples:
        .. code-block:: python

            x = layers.data(name='x', shape=[10])
            y = layers.data(name='y', shape=[10, 20], lod_level=2)
            out = layers.lod_reset(x=x, y=y)
    """
    helper = LayerHelper("lod_reset", **locals())
    out = helper.create_tmp_variable(dtype=x.dtype)
    if y is not None:
        helper.append_op(
            type="lod_reset", inputs={'X': x,
                                      'Y': y}, outputs={'Out': out})
    elif target_lod is not None:
        helper.append_op(
            type="lod_reset",
            inputs={'X': x},
            attrs={'target_lod': target_lod},
            outputs={'Out': out})
    else:
        raise ValueError("y and target_lod should not be both None.")

    return out
D
dragonwarrior 已提交
4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402


def lrn(input, n=5, k=1.0, alpha=1e-4, beta=0.75, name=None):
    """
    Local Response Normalization Layer. This layer performs a type of
    "lateral inhibition" by normalizing over local input regions.

    The formula is as follows:

    .. math::

D
dzhwinter 已提交
4403
      Output(i, x, y) = Input(i, x, y) / \\left(k + \\alpha \\sum\\limits^{\\min(C, c + n/2)}_{j = \\max(0, c - n/2)}(Input(j, x, y))^2\\right)^{\\beta}
D
dragonwarrior 已提交
4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431

    In the above equation:

    * :math:`n`: The number of channels to sum over.
    * :math:`k`: The offset (avoid being divided by 0).
    * :math:`alpha`: The scaling parameter.
    * :math:`beta`: The exponent parameter.

    Refer to `ImageNet Classification with Deep Convolutional Neural Networks
    <https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf>`_

    Args:
        input (Variable): The input tensor of this layer, and the dimension of input tensor must be 4.
        n (int, default 5): The number of channels to sum over.
        k (float, default 1.0): An offset (usually positive to avoid dividing by 0).
        alpha (float, default 1e-4): The scaling parameter.
        beta (float, default 0.75): The exponent.
        name (str, default None): A name for this operation.

    Raises:
        ValueError: If rank of the input tensor is not 4.

    Returns:
        A tensor variable storing the transformation result.

    Examples:
        .. code-block:: python

F
stash  
fengjiayi 已提交
4432 4433
          data = fluid.layers.data(
              name="data", shape=[3, 112, 112], dtype="float32")
D
dragonwarrior 已提交
4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460
          lrn = fluid.layers.lrn(input=data)
    """
    helper = LayerHelper('lrn', **locals())
    dtype = helper.input_dtype()
    input_shape = input.shape
    dims = len(input_shape)

    if dims != 4:
        raise ValueError(
            "dims of input must be 4(not %d), and it's order must be NCHW" %
            (dims))

    mid_out = helper.create_tmp_variable(dtype=dtype, stop_gradient=True)
    lrn_out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type="lrn",
        inputs={"X": input},
        outputs={
            "Out": lrn_out,
            "MidOut": mid_out,
        },
        attrs={"n": n,
               "k": k,
               "alpha": alpha,
               "beta": beta})

    return lrn_out
G
guosheng 已提交
4461 4462 4463 4464


def pad(x, paddings, pad_value=0., name=None):
    """
G
guosheng 已提交
4465
    Pads a tensor with a constant value given by :attr:`pad_value`, and the
W
wanghaoshuang 已提交
4466
    padded width is specified by :attr:`paddings`.
G
guosheng 已提交
4467

G
guosheng 已提交
4468 4469 4470 4471
    Specifically, the number of values padded before the contents of :attr:`x`
    in dimension :attr:`i` is indicated by :attr:`paddings[i]`, and the number
    of values padded after the contents of :attr:`x` in dimension :attr:`i` is
    indicated by :attr:`paddings[i+1]`.
G
guosheng 已提交
4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493

    See below for an example.

    .. code-block:: text

        Given:
            x = [[1, 2], [3, 4]]

            paddings = [0, 1, 1, 2]

            pad_value = 0

        Return:

            out = [[0, 1, 2, 0, 0]
                   [0, 3, 4, 0, 0]
                   [0, 0, 0, 0, 0]]

    Args:
        x (Variable): The input tensor variable.
        paddings (list): A list of integers. Its elements specify the padded
                         width before and after for each dimension in turn.
W
wanghaoshuang 已提交
4494
                         The length of :attr:paddings must be
G
guosheng 已提交
4495 4496 4497 4498 4499 4500 4501 4502 4503 4504
                         :math:`rank(x) \\times 2`.
        pad_value (float): The constant value used to pad.
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.

    Returns:
        Variable: The padded tensor variable.

    Examples:
        .. code-block:: python
G
guosheng 已提交
4505

G
guosheng 已提交
4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519
            # x is a rank 2 tensor variable.
            out = fluid.layers.pad(
                x=x, paddings=[0, 1, 1, 2], pad_value=0.)
    """
    helper = LayerHelper('pad', input=x, **locals())
    dtype = helper.input_dtype()
    out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type='pad',
        inputs={'X': x},
        outputs={'Out': out},
        attrs={'paddings': paddings,
               'pad_value': float(pad_value)})
    return out
4520 4521 4522 4523 4524 4525 4526 4527 4528


def label_smooth(label,
                 prior_dist=None,
                 epsilon=0.1,
                 dtype="float32",
                 name=None):
    """
    Label smoothing is a mechanism to regularize the classifier layer and is
4529 4530
    called label-smoothing regularization (LSR).

4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553
    Label smoothing is proposed to encourage the model to be less confident,
    since optimizing the log-likelihood of the correct label directly may
    cause overfitting and reduce the ability of the model to adapt. Label
    smoothing replaces the ground-truth label :math:`y` with the weighted sum
    of itself and some fixed distribution :math:`\mu`. For class :math:`k`,
    i.e.

    .. math::

        \\tilde{y_k} = (1 - \epsilon) * y_k + \epsilon * \mu_k,

    where :math:`1 - \epsilon` and :math:`\epsilon` are the weights
    respectively, and :math:`\\tilde{y}_k` is the smoothed label. Usually
    uniform distribution is used for :math:`\mu`.

    See more details about label smoothing in https://arxiv.org/abs/1512.00567.

    Args:
        label(Variable): The input variable containing the label data. The
                          label data should use one-hot representation.
        prior_dist(Variable): The prior distribution to be used to smooth
                              labels. If not provided, an uniform distribution
                              is used. The shape of :attr:`prior_dist` should
4554
                              be :math:`(1, class\_num)`.
4555 4556
        epsilon(float): The weight used to mix up the original ground-truth
                        distribution and the fixed distribution.
4557
        dtype(np.dtype|core.VarDesc.VarType|str): The type of data : float32,
4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584
                                                  float_64, int etc.
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.

    Returns:
        Variable: The tensor variable containing the smoothed labels.

    Examples:
        .. code-block:: python

            label = layers.data(name="label", shape=[1], dtype="float32")
            one_hot_label = layers.one_hot(input=label, depth=10)
            smooth_label = layers.label_smooth(
                label=one_hot_label, epsilon=0.1, dtype="float32")
    """
    if epsilon > 1. or epsilon < 0.:
        raise ValueError("The value of epsilon must be between 0 and 1.")
    helper = LayerHelper("label_smooth", **locals())
    label.stop_gradient = True
    smooth_label = helper.create_tmp_variable(dtype)
    helper.append_op(
        type="label_smooth",
        inputs={"X": label,
                "PriorDist": prior_dist} if prior_dist else {"X": label},
        outputs={"Out": smooth_label},
        attrs={"epsilon": float(epsilon)})
    return smooth_label
4585 4586


Y
yi.wu 已提交
4587
@templatedoc()
4588 4589
def roi_pool(input, rois, pooled_height=1, pooled_width=1, spatial_scale=1.0):
    """
Y
yi.wu 已提交
4590
    ${comment}
4591 4592

    Args:
Y
yi.wu 已提交
4593 4594
        input (Variable): ${x_comment}
        rois (Variable): ROIs (Regions of Interest) to pool over.
Y
yi.wu 已提交
4595 4596 4597
        pooled_height (integer): ${pooled_height_comment} Default: 1
        pooled_width (integer): ${pooled_width_comment} Default: 1
        spatial_scale (float): ${spatial_scale_comment} Default: 1.0
4598 4599

    Returns:
Y
update  
yi.wu 已提交
4600
        Variable: ${out_comment}.
4601 4602

    Examples:
4603 4604
        .. code-block:: python

4605
            pool_out = fluid.layers.roi_pool(input=x, rois=rois, 7, 7, 1.0)
4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622
    """
    helper = LayerHelper('roi_pool', **locals())
    dtype = helper.input_dtype()
    pool_out = helper.create_tmp_variable(dtype)
    argmaxes = helper.create_tmp_variable(dtype='int32')
    helper.append_op(
        type="roi_pool",
        inputs={"X": input,
                "ROIs": rois},
        outputs={"Out": pool_out,
                 "Argmax": argmaxes},
        attrs={
            "pooled_height": pooled_height,
            "pooled_width": pooled_width,
            "spatial_scale": spatial_scale
        })
    return pool_out
W
whs 已提交
4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650


def dice_loss(input, label, epsilon=0.00001):
    """
    Dice loss for comparing the similarity of two batch of data,
    usually is used for binary image segmentation i.e. labels are binary.
    The dice loss can be defined as below equation:

    .. math::

        dice\_loss &= 1 - \\frac{2 * intersection\_area}{total\_area} \\\\
                  &= \\frac{(total\_area - intersection\_area) - intersection\_area}{total\_area} \\\\
                  &= \\frac{(union\_area - intersection\_area)}{total\_area}


    Args:
        input (Variable): The predictions with rank>=2. The first dimension is batch size,
                          and the last dimension is class number.
        label (Variable): The groud truth with the same rank with input. The first dimension
                          is batch size, and the last dimension is 1.
        epsilon (float): The epsilon will be added to the numerator and denominator.
                         If both input and label are empty, it makes sure dice is 1.
                         Default: 0.00001

    Returns:
        dice_loss (Variable): The dice loss with shape [1].

    Examples:
4651 4652
        .. code-block:: python

W
whs 已提交
4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663
            predictions = fluid.layers.softmax(x)
            loss = fluid.layers.dice_loss(input=predictions, label=label, 2)
    """
    label = one_hot(label, depth=input.shape[-1])
    reduce_dim = range(1, len(input.shape))
    inse = reduce_sum(input * label, dim=reduce_dim)
    dice_denominator = reduce_sum(
        input, dim=reduce_dim) + reduce_sum(
            label, dim=reduce_dim)
    dice_score = 1 - inse * 2 / (dice_denominator + epsilon)
    return reduce_mean(dice_score)
4664 4665


4666 4667 4668 4669 4670
def image_resize(input,
                 out_shape=None,
                 scale=None,
                 name=None,
                 resample='BILINEAR'):
4671
    """
Q
qiaolongfei 已提交
4672
    **Resize a Batch of Images**
F
stash  
fengjiayi 已提交
4673

4674 4675 4676 4677
    The input must be a tensor of the shape (num_batches, channels, in_h, in_w), 
    and the resizing only applies on the last two dimensions(hight and width).

    Supporting resample methods:
Q
update  
qiaolongfei 已提交
4678

4679
        'BILINEAR' : Bilinear interpolation
F
stash  
fengjiayi 已提交
4680

4681
    Args:
4682
        input (Variable): The input tensor of image resize layer,
4683 4684
                          This is a 4-D tensor of the shape
                          (num_batches, channels, in_h, in_w).
4685
        out_shape(list|tuple|Variable|None): Output shape of image resize
4686 4687
                                    layer, the shape is (out_h, out_w).
                                    Default: None
B
baiyf 已提交
4688
        scale(float|None): The multiplier for the input height or width.
4689 4690 4691
                         At least one of out_shape or scale must be set.
                         And out_shape has a higher priority than scale.
                         Default: None
4692 4693
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
4694 4695
        resample(str): The resample method. It can only be 'BILINEAR' currently.
                       Default: 'BILINEAR'
4696 4697

    Returns:
Q
update  
qiaolongfei 已提交
4698 4699
        Variable: The output is a 4-D tensor of the shape
        (num_batches, channls, out_h, out_w).
F
stash  
fengjiayi 已提交
4700

4701 4702 4703
    Examples:
        .. code-block:: python

4704
            out = fluid.layers.image_resize(input, out_shape=[12, 12])
4705
    """
4706 4707 4708 4709
    resample_methods = {'BILINEAR': 'bilinear_interp'}
    if resample not in resample_methods:
        raise ValueError(
            "The 'resample' of image_resize can only be 'BILINEAR' currently.")
4710 4711
    if out_shape is None and scale is None:
        raise ValueError("One of out_shape and scale must not be None")
4712 4713
    helper = LayerHelper('bilinear_interp', **locals())
    dtype = helper.input_dtype()
4714 4715 4716 4717

    def _is_list_or_turple_(data):
        return (isinstance(data, list) or isinstance(data, tuple))

4718 4719 4720
    out_h = 0
    out_w = 0
    inputs = {"X": input}
4721
    if out_shape is not None:
B
baiyf 已提交
4722 4723 4724
        if not (_is_list_or_turple_(out_shape) and
                len(out_shape) == 2) and not isinstance(out_shape, Variable):
            raise ValueError('out_shape should be a list or tuple or variable')
4725 4726 4727 4728 4729 4730
        if _is_list_or_turple_(out_shape):
            out_shape = list(map(int, out_shape))
            out_h = out_shape[0]
            out_w = out_shape[1]
        else:
            inputs['OutSize'] = out_shape
4731 4732 4733 4734
    else:
        out_h = int(input.shape[2] * scale)
        out_w = int(input.shape[3] * scale)

4735 4736
    out = helper.create_tmp_variable(dtype)
    helper.append_op(
4737
        type=resample_methods[resample],
4738
        inputs=inputs,
4739 4740 4741 4742
        outputs={"Out": out},
        attrs={"out_h": out_h,
               "out_w": out_w})
    return out
F
stash  
fengjiayi 已提交
4743 4744


Y
yuyang18 已提交
4745
@templatedoc(op_type="bilinear_interp")
4746 4747
def resize_bilinear(input, out_shape=None, scale=None, name=None):
    """
Y
yuyang18 已提交
4748 4749 4750 4751 4752 4753
    ${comment}

    Args:
        input(${x_type}): ${x_comment}.

        out_shape(${out_size_type}): ${out_size_comment}.
4754

Y
yuyang18 已提交
4755 4756 4757 4758 4759 4760 4761 4762
        scale(float|None): The multiplier for the input height or width. At
             least one of out_shape or scale must be set. And out_shape has
             a higher priority than scale. Default: None.

        name(str|None): The output variable name.

    Returns:
        ${out_comment}.
4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779
    """

    return image_resize(input, out_shape, scale, name, 'BILINEAR')


def image_resize_short(input, out_short_len, resample='BILINEAR'):
    """
    Resize a batch of images. The short edge of input images will be 
    resized to the given 'out_short_len'. The long edge of input images 
    will be resized proportionately to make images' length-width ratio 
    constant.

    Args:
        input (Variable): The input tensor of image resize layer,
                          This is a 4-D tensor of the shape
                          (num_batches, channels, in_h, in_w).
        out_short_len(int): The length of output images' short edge.
4780
        resample (str): resample method, default: BILINEAR.
F
fengjiayi 已提交
4781

4782
    Returns:
Q
update  
qiaolongfei 已提交
4783
        Variable: The output is a 4-D tensor of the shape
4784
        (num_batches, channls, out_h, out_w).
4785 4786 4787 4788 4789 4790 4791 4792 4793 4794
    """
    in_shape = input.shape
    if len(in_shape) != 4:
        raise ValueError(
            "The rank of input must be 4 (num_batches, channels, in_h, in_w).")
    hw = in_shape[2:4]
    short_idx = hw.index(min(hw))
    long_idx = 1 - short_idx
    out_shape = list(hw)
    out_shape[short_idx] = out_short_len
F
fengjiayi 已提交
4795 4796 4797
    out_shape[long_idx] = int(
        float(out_shape[long_idx]) * (float(out_short_len) / float(hw[
            short_idx])) + 0.5)
4798 4799 4800
    return image_resize(input=input, out_shape=out_shape, resample=resample)


W
whs 已提交
4801 4802
def gather(input, index):
    """
Q
qiaolongfei 已提交
4803 4804
    **Gather Layer**

W
whs 已提交
4805 4806 4807 4808 4809
    Output is obtained by gathering entries of the outer-most dimension 
    of X indexed by `index` and concatenate them together.

    .. math::

4810
        Out = X[Index]
W
whs 已提交
4811 4812 4813 4814 4815 4816 4817


    .. code-block:: text


                Given:

4818 4819
                X = [[1, 2],
                     [3, 4],
W
whs 已提交
4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
                     [5, 6]]

                Index = [1, 2]

                Then:

                Out = [[3, 4],
                       [5, 6]]

    Args:
        input (Variable): The source input with rank>=1. 
        index (Variable): The index input with rank=1.

    Returns:
        output (Variable): The output is a tensor with the same rank as input.

    Examples:
W
whs 已提交
4837

W
whs 已提交
4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852
        .. code-block:: python

            output = fluid.layers.gather(x, index)
    """
    helper = LayerHelper('gather', **locals())
    dtype = helper.input_dtype()
    out = helper.create_tmp_variable(dtype)
    helper.append_op(
        type="gather",
        inputs={"X": input,
                "Index": index},
        outputs={"Out": out})
    return out


Y
yuyang18 已提交
4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865
@templatedoc()
def random_crop(x, shape, seed=None):
    """
    ${comment}

    Args:
        x(${x_type}): ${x_comment}
        shape(${shape_type}): ${shape_comment}
        seed(int|${seed_type}|None): ${seed_comment} By default, the seed will
            get from `random.randint(-65536, 65535)`.

    Returns:
        ${out_comment}
F
fengjiayi 已提交
4866

4867 4868 4869
    Examples:
        >>> img = fluid.layers.data("img", [3, 256, 256])
        >>> cropped_img = fluid.layers.random_crop(img, shape=[3, 224, 224])
Y
yuyang18 已提交
4870
    """
F
stash  
fengjiayi 已提交
4871
    helper = LayerHelper("random_crop", **locals())
F
fengjiayi 已提交
4872
    dtype = x.dtype
F
stash  
fengjiayi 已提交
4873
    out = helper.create_tmp_variable(dtype)
Y
yuyang18 已提交
4874 4875
    if seed is None:
        seed = random.randint(-65536, 65535)
F
fengjiayi 已提交
4876
    op_attrs = {"shape": shape}
F
stash  
fengjiayi 已提交
4877
    if isinstance(seed, int):
F
fengjiayi 已提交
4878 4879 4880 4881 4882
        op_attrs["startup_seed"] = seed
        seed = helper.create_variable(
            name=unique_name.generate("random_crop_seed"),
            dtype="int64",
            persistable=True)
F
stash  
fengjiayi 已提交
4883 4884 4885 4886
    elif not isinstance(seed, Variable):
        raise ValueError("'seed' must be a Variable or an int.")
    helper.append_op(
        type="random_crop",
F
fix  
fengjiayi 已提交
4887
        inputs={"X": x,
F
stash  
fengjiayi 已提交
4888 4889
                "Seed": seed},
        outputs={"Out": out,
F
fengjiayi 已提交
4890 4891
                 "SeedOut": seed},
        attrs=op_attrs)
F
stash  
fengjiayi 已提交
4892
    return out
W
whs 已提交
4893 4894


W
wanghaoshuang 已提交
4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948
def log(x):
    """
    Calculates the natural log of the given input tensor, element-wise.

    .. math::

        Out = \\ln(x)

    Args:
        x (Variable): Input tensor. 

    Returns:
        Variable: The natural log of the input tensor computed element-wise.

    Examples:

        .. code-block:: python

            output = fluid.layers.log(x)
    """
    helper = LayerHelper('log', **locals())
    dtype = helper.input_dtype()
    out = helper.create_tmp_variable(dtype)
    helper.append_op(type="log", inputs={"X": input}, outputs={"Out": out})
    return out


def relu(x):
    """
    Relu takes one input data (Tensor) and produces one output data (Tensor)
    where the rectified linear function, y = max(0, x), is applied to
    the tensor elementwise.

    .. math::

        Out = \\max(0, x)

    Args:
        x (Variable): The input tensor. 

    Returns:
        Variable: The output tensor with the same shape as input.

    Examples:

        .. code-block:: python

            output = fluid.layers.relu(x)
    """
    helper = LayerHelper('relu', **locals())
    dtype = helper.input_dtype()
    out = helper.create_tmp_variable(dtype)
    helper.append_op(type="relu", inputs={"X": input}, outputs={"Out": out})
    return out
4949 4950


W
whs 已提交
4951 4952 4953 4954 4955 4956
def mean_iou(input, label, num_classes):
    """
    Mean Intersection-Over-Union is a common evaluation metric for
    semantic image segmentation, which first computes the IOU for each 
    semantic class and then computes the average over classes. 
    IOU is defined as follows: 
F
fengjiayi 已提交
4957

W
whs 已提交
4958
    .. math::
4959 4960

        IOU = \\frac{true\_positiv}{(true\_positive + false\_positive + false\_negative)}.
W
whs 已提交
4961 4962 4963 4964 4965 4966 4967

    The predictions are accumulated in a confusion matrix and mean-IOU 
    is then calculated from it.


    Args:
        input (Variable): A Tensor of prediction results for semantic labels with type int32 or int64.
4968
        label (Variable): A Tensor of ground truth labels with type int32 or int64.
W
whs 已提交
4969
                           Its shape should be the same as input.
4970
        num_classes (int): The possible number of labels.
W
whs 已提交
4971 4972 4973 4974 4975 4976 4977 4978 4979

    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. 

    Examples:

        .. code-block:: python
F
fengjiayi 已提交
4980

W
whs 已提交
4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998
            iou, wrongs, corrects = fluid.layers.mean_iou(predict, label, num_classes)
    """
    helper = LayerHelper('mean_iou', **locals())
    dtype = helper.input_dtype()
    out_mean_iou = helper.create_tmp_variable(dtype='float32')
    out_wrong = helper.create_tmp_variable(dtype='int32')
    out_correct = helper.create_tmp_variable(dtype='int32')
    helper.append_op(
        type="mean_iou",
        inputs={"predictions": input,
                "labels": label},
        outputs={
            "out_mean_iou": out_mean_iou,
            "out_wrong": out_wrong,
            "out_correct": out_correct
        },
        attrs={"num_classes": num_classes})
    return out_mean_iou, out_wrong, out_correct