nn.py 153.0 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 16 17 18 19 20
"""
All layers just related to the neural network.
"""

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
Y
Yu Yang 已提交
26 27

__all__ = [
Y
ying 已提交
28 29 30
    'fc',
    'embedding',
    'dynamic_lstm',
Y
Yibing Liu 已提交
31
    'dynamic_lstmp',
G
guosheng 已提交
32
    'dynamic_gru',
Y
ying 已提交
33 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',
    'sequence_pool',
43 44
    'sequence_softmax',
    'softmax',
Y
ying 已提交
45 46 47 48 49 50 51 52 53 54
    'pool2d',
    'batch_norm',
    'beam_search_decode',
    'conv2d_transpose',
    'sequence_expand',
    'lstm_unit',
    'reduce_sum',
    'reduce_mean',
    'reduce_max',
    'reduce_min',
55
    'reduce_prod',
Y
ying 已提交
56 57 58 59
    'sequence_first_step',
    'sequence_last_step',
    'dropout',
    'split',
60 61
    'ctc_greedy_decoder',
    'edit_distance',
Y
ying 已提交
62 63
    'l2_normalize',
    'matmul',
Q
qingqing01 已提交
64
    'topk',
Y
ying 已提交
65 66
    'warpctc',
    'sequence_reshape',
67
    'transpose',
68
    'im2sequence',
69
    'nce',
Q
Qiao Longfei 已提交
70
    'beam_search',
71
    'row_conv',
72
    'multiplex',
G
guosheng 已提交
73
    'layer_norm',
74 75
    'softmax_with_cross_entropy',
    'smooth_l1',
76
    'one_hot',
Y
Yu Yang 已提交
77
    'autoincreased_step_counter',
C
caoying03 已提交
78
    'reshape',
Y
yangyaming 已提交
79
    'lod_reset',
D
dragonwarrior 已提交
80
    'lrn',
G
guosheng 已提交
81
    'pad',
82
    'label_smooth',
83
    'roi_pool',
W
whs 已提交
84
    'dice_loss',
F
fengjiayi 已提交
85 86
    'image_resize',
    'image_resize_short',
B
baiyf 已提交
87
    'resize_bilinear',
W
whs 已提交
88
    'gather',
89
    'random_crop',
Y
Yu Yang 已提交
90 91 92 93 94 95 96 97
]


def fc(input,
       size,
       num_flatten_dims=1,
       param_attr=None,
       bias_attr=None,
K
Kexin Zhao 已提交
98
       use_cudnn=False,
99
       use_mkldnn=False,
Y
Yu Yang 已提交
100
       act=None,
J
Jacek Czaja 已提交
101
       is_test=False,
102
       name=None):
Y
Yu Yang 已提交
103
    """
104
    **Fully Connected Layer**
Y
Yu Yang 已提交
105

C
caoying03 已提交
106
    The fully connected layer can take multiple tensors as its inputs. It
R
ranqiu 已提交
107 108 109 110 111 112
    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,
Y
ying 已提交
113
    if activation is not None, it will be applied to the output as well.
C
caoying03 已提交
114

C
caoying03 已提交
115
    This process can be formulated as follows:
116 117 118

    .. math::

119
        Out = Act({\sum_{i=0}^{N-1}X_iW_i + b})
120 121 122

    In the above equation:

C
caoying03 已提交
123 124 125 126
    * :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).
127
    * :math:`Act`: The activation function.
C
caoying03 已提交
128
    * :math:`Out`: The output tensor.
Y
Yu Yang 已提交
129 130

    Args:
R
ranqiu 已提交
131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
        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
148
        is_test(bool): A flag indicating whether execution is in test phase.
M
mozga-intel 已提交
149 150
        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
151
        name (str, default None): The name of this layer.
Y
Yu Yang 已提交
152

153
    Returns:
R
ranqiu 已提交
154
        A tensor variable storing the transformation result.
155 156

    Raises:
C
caoying03 已提交
157
        ValueError: If rank of the input tensor is less than 2.
158 159 160 161

    Examples:
        .. code-block:: python

F
stash  
fengjiayi 已提交
162 163
          data = fluid.layers.data(
              name="data", shape=[32, 32], dtype="float32")
164
          fc = fluid.layers.fc(input=data, size=1000, act="tanh")
Y
Yu Yang 已提交
165
    """
C
caoying03 已提交
166

C
caoying03 已提交
167
    helper = LayerHelper("fc", **locals())
Y
Yu Yang 已提交
168 169 170 171

    dtype = helper.input_dtype()

    mul_results = []
172 173
    for input_var, param_attr in helper.iter_inputs_and_params():
        input_shape = input_var.shape
Y
Yu Yang 已提交
174 175 176
        param_shape = [
            reduce(lambda a, b: a * b, input_shape[num_flatten_dims:], 1)
        ] + [size]
Y
ying 已提交
177

Y
Yu Yang 已提交
178
        w = helper.create_parameter(
179 180
            attr=param_attr, shape=param_shape, dtype=dtype, is_bias=False)
        tmp = helper.create_tmp_variable(dtype)
181
        helper.append_op(
182 183 184
            type="mul",
            inputs={"X": input_var,
                    "Y": w},
185
            outputs={"Out": tmp},
M
mozga-intel 已提交
186 187
            attrs={"x_num_col_dims": num_flatten_dims,
                   "y_num_col_dims": 1})
188 189 190 191
        mul_results.append(tmp)

    if len(mul_results) == 1:
        pre_bias = mul_results[0]
192
    else:
193 194 195 196 197 198 199
        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
200 201


202 203 204
def embedding(input,
              size,
              is_sparse=False,
205
              is_distributed=False,
206 207 208
              padding_idx=None,
              param_attr=None,
              dtype='float32'):
Y
Yu Yang 已提交
209
    """
210 211
    **Embedding Layer**

212
    This layer is used to lookup embeddings of IDs, provided by :attr:`input`, in
213 214
    a lookup table. The result of this lookup is the embedding of each ID in the
    :attr:`input`.
215 216 217

    All the input variables are passed in as local variables to the LayerHelper
    constructor.
Y
Yu Yang 已提交
218 219

    Args:
220 221 222 223 224 225 226
        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.
        padding_idx(int|long|None): If :attr:`None`, it makes no effect to lookup.
            Otherwise the given :attr:`padding_idx` indicates padding the output
227 228
            with zeros whenever lookup encounters it in :attr:`input`. If
            :math:`padding_idx < 0`, the padding_idx to use in lookup is
229 230
            :math:`size[0] + dim`.
        param_attr(ParamAttr): Parameters for this layer
231
        dtype(np.dtype|core.VarDesc.VarType|str): The type of data : float32, float_16, int etc
Y
Yu Yang 已提交
232

233 234 235
    Returns:
        Variable: The tensor variable storing the embeddings of the \
                  supplied inputs.
Y
Yu Yang 已提交
236

237 238
    Examples:
        .. code-block:: python
Y
Yu Yang 已提交
239

C
chengduoZH 已提交
240
          dict_size = len(dataset.ids)
241
          data = fluid.layers.data(name='ids', shape=[32, 32], dtype='float32')
C
chengduoZH 已提交
242
          fc = fluid.layers.embedding(input=data, size=[dict_size, 16])
Y
Yu Yang 已提交
243 244 245 246 247 248
    """

    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)
249 250
    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
251 252 253 254 255
    helper.append_op(
        type='lookup_table',
        inputs={'Ids': input,
                'W': w},
        outputs={'Out': tmp},
256 257 258 259 260
        attrs={
            'is_sparse': is_sparse,
            'is_distributed': is_distributed,
            'padding_idx': padding_idx
        })
Y
Yu Yang 已提交
261 262 263 264 265 266 267 268 269 270 271 272 273
    return tmp


# TODO(qijun): expose H0 and C0
def dynamic_lstm(input,
                 size,
                 param_attr=None,
                 bias_attr=None,
                 use_peepholes=True,
                 is_reverse=False,
                 gate_activation='sigmoid',
                 cell_activation='tanh',
                 candidate_activation='tanh',
274 275
                 dtype='float32',
                 name=None):
Y
Yibing Liu 已提交
276 277 278 279 280 281
    """
    **Dynamic LSTM Layer**

    The defalut implementation is diagonal/peephole connection
    (https://arxiv.org/pdf/1402.1128.pdf), the formula is as follows:

Y
Yibing Liu 已提交
282
    .. math::
Y
Yibing Liu 已提交
283

284
        i_t & = \sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + W_{ic}c_{t-1} + b_i)
Y
Yibing Liu 已提交
285

286
        f_t & = \sigma(W_{fx}x_{t} + W_{fh}h_{t-1} + W_{fc}c_{t-1} + b_f)
Y
Yibing Liu 已提交
287

288
        \\tilde{c_t} & = act_g(W_{cx}x_t + W_{ch}h_{t-1} + b_c)
Y
Yibing Liu 已提交
289

290 291 292
        o_t & = \sigma(W_{ox}x_{t} + W_{oh}h_{t-1} + W_{oc}c_t + b_o)

        c_t & = f_t \odot c_{t-1} + i_t \odot \\tilde{c_t}
Y
Yibing Liu 已提交
293

Y
Yibing Liu 已提交
294
        h_t & = o_t \odot act_h(c_t)
Y
Yibing Liu 已提交
295

296
    where the :math:`W` terms denote weight matrices (e.g. :math:`W_{xi}` is
297
    the matrix of weights from the input gate to the input), :math:`W_{ic}, \
298 299 300
    W_{fc}, W_{oc}` are diagonal weight matrices for peephole connections. In
    our implementation, we use vectors to reprenset these diagonal weight
    matrices. The :math:`b` terms denote bias vectors (:math:`b_i` is the input
Y
Yibing Liu 已提交
301
    gate bias vector), :math:`\sigma` is the non-linear activations, such as
302 303
    logistic sigmoid function, and :math:`i, f, o` and :math:`c` are the input
    gate, forget gate, output gate, and cell activation vectors, respectively,
304 305
    all of which have the same size as the cell output activation vector :math:`h`.

306 307 308 309
    The :math:`\odot` is the element-wise product of the vectors. :math:`act_g`
    and :math:`act_h` are the cell input and cell output activation functions
    and `tanh` is usually used for them. :math:`\\tilde{c_t}` is also called
    candidate hidden state, which is computed based on the current input and
310 311 312
    the previous hidden state.

    Set `use_peepholes` to `False` to disable peephole connection. The formula
Y
Yibing Liu 已提交
313 314 315
    is omitted here, please refer to the paper
    http://www.bioinf.jku.at/publications/older/2604.pdf for details.

Y
Yibing Liu 已提交
316 317 318
    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-connect layer before LSTM layer.
Y
Yibing Liu 已提交
319 320

    Args:
321 322 323 324
        input(Variable): The input of dynamic_lstm 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
Y
Yibing Liu 已提交
325 326
                         mini-batch, D is the hidden size.
        size(int): 4 * hidden size.
327
        param_attr(ParamAttr|None): The parameter attribute for the learnable
328
                               hidden-hidden weights.
Y
Yibing Liu 已提交
329 330 331

                               - Weights = {:math:`W_{ch}, W_{ih}, \
                                                W_{fh}, W_{oh}`}
332 333 334
                               - The shape is (D x 4D), where D is the hidden
                                 size.
        bias_attr(ParamAttr|None): The bias attribute for the learnable bias
335 336 337
                              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
338

339
                              1. `use_peepholes = False`
Y
Yibing Liu 已提交
340
                                - Biases = {:math:`b_c, b_i, b_f, b_o`}.
341
                                - The shape is (1 x 4D).
342
                              2. `use_peepholes = True`
Y
Yibing Liu 已提交
343 344
                                - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \
                                                 W_{fc}, W_{oc}`}.
345
                                - The shape is (1 x 7D).
346
        use_peepholes(bool): Whether to enable diagonal/peephole connections,
Y
Yibing Liu 已提交
347 348
                             default `True`.
        is_reverse(bool): Whether to compute reversed LSTM, default `False`.
349 350
        gate_activation(str): The activation for input gate, forget gate and
                              output gate. Choices = ["sigmoid", "tanh", "relu",
Y
Yibing Liu 已提交
351
                              "identity"], default "sigmoid".
352
        cell_activation(str): The activation for cell output. Choices = ["sigmoid",
Y
Yibing Liu 已提交
353 354
                              "tanh", "relu", "identity"], default "tanh".
        candidate_activation(str): The activation for candidate hidden state.
F
stash  
fengjiayi 已提交
355 356
                              Choices = ["sigmoid", "tanh",
                                  "relu", "identity"],
Y
Yibing Liu 已提交
357 358
                              default "tanh".
        dtype(str): Data type. Choices = ["float32", "float64"], default "float32".
359 360
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
Y
Yibing Liu 已提交
361 362

    Returns:
Y
Yibing Liu 已提交
363 364
        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
365

Y
Yibing Liu 已提交
366
    Examples:
Y
Yibing Liu 已提交
367 368
        .. code-block:: python

Y
Yibing Liu 已提交
369 370
            hidden_dim = 512
            forward_proj = fluid.layers.fc(input=input_seq, size=hidden_dim * 4,
371
                                           act=None, bias_attr=None)
Y
Yibing Liu 已提交
372 373
            forward, _ = fluid.layers.dynamic_lstm(
                input=forward_proj, size=hidden_dim * 4, use_peepholes=False)
Y
Yibing Liu 已提交
374
    """
375

Y
Yu Yang 已提交
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411
    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)

    helper.append_op(
        type='lstm',
        inputs={'Input': input,
                'Weight': weight,
                'Bias': bias},
        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
412 413 414 415 416 417 418 419 420 421 422
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',
423 424
                  dtype='float32',
                  name=None):
Y
Yibing Liu 已提交
425 426 427
    """
    **Dynamic LSTMP Layer**

428 429 430 431 432 433
    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
434 435 436 437 438

    The formula is as follows:

    .. math::

439
        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
440

441
        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
442

443
        \\tilde{c_t} & = act_g(W_{cx}x_t + W_{cr}r_{t-1} + b_c)
Y
Yibing Liu 已提交
444

445
        o_t & = \sigma(W_{ox}x_{t} + W_{or}r_{t-1} + W_{oc}c_t + b_o)
Y
Yibing Liu 已提交
446

447
        c_t & = f_t \odot c_{t-1} + i_t \odot \\tilde{c_t}
Y
Yibing Liu 已提交
448

449
        h_t & = o_t \odot act_h(c_t)
Y
Yibing Liu 已提交
450

451
        r_t & = \overline{act_h}(W_{rh}h_t)
Y
Yibing Liu 已提交
452

Y
Yibing Liu 已提交
453 454 455 456 457 458
    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, \
459
          we use vectors to reprenset these diagonal weight matrices.
Y
Yibing Liu 已提交
460
    * :math:`b`: Denotes bias vectors (e.g. :math:`b_i` is the input gate \
461
          bias vector).
Y
Yibing Liu 已提交
462 463 464
    * :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 \
465
          the same size as the cell output activation vector :math:`h`.
Y
Yibing Liu 已提交
466
    * :math:`h`: The hidden state.
467
    * :math:`r`: The recurrent projection of the hidden state.
Y
Yibing Liu 已提交
468 469
    * :math:`\\tilde{c_t}`: The candidate hidden state, whose \
          computation is based on the current input and previous hidden state.
470
    * :math:`\odot`: The element-wise product of the vectors.
Y
Yibing Liu 已提交
471
    * :math:`act_g` and :math:`act_h`: The cell input and cell output \
472
          activation functions and `tanh` is usually used for them.
Y
Yibing Liu 已提交
473 474
    * :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
475 476 477 478

    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.
479

Y
Yibing Liu 已提交
480 481 482 483 484 485 486 487 488 489 490 491
    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.
492
        param_attr(ParamAttr|None): The parameter attribute for the learnable
Y
Yibing Liu 已提交
493 494
                               hidden-hidden weight and projection weight.

495 496
                               - Hidden-hidden weight = {:math:`W_{ch}, W_{ih}, \
                                                W_{fh}, W_{oh}`}.
497 498
                               - 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
499 500
                                 size.
                               - Projection weight = {:math:`W_{rh}`}.
501 502
                               - 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
503 504 505 506 507 508
                              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`}.
509
                                - The shape is (1 x 4D).
Y
Yibing Liu 已提交
510 511 512
                              2. `use_peepholes = True`
                                - Biases = { :math:`b_c, b_i, b_f, b_o, W_{ic}, \
                                                 W_{fc}, W_{oc}`}.
513
                                - The shape is (1 x 7D).
Y
Yibing Liu 已提交
514 515 516 517 518 519 520 521 522
        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.
F
stash  
fengjiayi 已提交
523 524
                              Choices = ["sigmoid", "tanh",
                                  "relu", "identity"],
Y
Yibing Liu 已提交
525 526
                              default "tanh".
        proj_activation(str): The activation for projection output.
F
stash  
fengjiayi 已提交
527 528
                              Choices = ["sigmoid", "tanh",
                                  "relu", "identity"],
Y
Yibing Liu 已提交
529 530
                              default "tanh".
        dtype(str): Data type. Choices = ["float32", "float64"], default "float32".
531 532
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
Y
Yibing Liu 已提交
533 534

    Returns:
535 536
        tuple: The projection of hidden state, and cell state of LSTMP. The \
               shape of projection is (T x P), for the cell state which is \
Y
Yibing Liu 已提交
537 538 539 540 541
               (T x D), and both LoD is the same with the `input`.

    Examples:
        .. code-block:: python

Y
Yibing Liu 已提交
542
            hidden_dim, proj_dim = 512, 256
Y
Yibing Liu 已提交
543 544
            fc_out = fluid.layers.fc(input=input_seq, size=hidden_dim * 4,
                                     act=None, bias_attr=None)
545 546 547
            proj_out, _ = fluid.layers.dynamic_lstmp(input=fc_out,
                                                     size=hidden_dim * 4,
                                                     proj_size=proj_dim,
Y
Yibing Liu 已提交
548 549 550 551
                                                     use_peepholes=False,
                                                     is_reverse=True,
                                                     cell_activation="tanh",
                                                     proj_activation="tanh")
Y
Yibing Liu 已提交
552
    """
553

Y
Yibing Liu 已提交
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599
    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
600 601 602 603 604 605 606 607 608 609 610
def dynamic_gru(input,
                size,
                param_attr=None,
                bias_attr=None,
                is_reverse=False,
                gate_activation='sigmoid',
                candidate_activation='tanh',
                h_0=None):
    """
    **Dynamic GRU Layer**

611
    Refer to `Empirical Evaluation of Gated Recurrent Neural Networks on
G
guosheng 已提交
612
    Sequence Modeling <https://arxiv.org/abs/1412.3555>`_
613

G
guosheng 已提交
614 615 616 617 618 619 620 621 622
    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)
623

G
guosheng 已提交
624
        h_t & = (1-u_t) \odot h_{t-1} + u_t \odot \\tilde{h_t}
625

G
guosheng 已提交
626
    The :math:`\odot` is the element-wise product of the vectors. :math:`act_g`
627 628
    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
629 630 631 632
    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
633
    to use fully-connect layer before GRU layer.
G
guosheng 已提交
634 635

    Args:
636 637
        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
638
            Variable is a matrix with shape :math:`(T \\times 3D)`, where
639
            :math:`T` is the total time steps in this mini-batch, :math:`D`
G
guosheng 已提交
640 641
            is the hidden size.
        size(int): The dimension of the gru cell.
642
        param_attr(ParamAttr|None): The parameter attribute for the learnable
G
guosheng 已提交
643 644
            hidden-hidden weight matrix. Note:

645
            - The shape of the weight matrix is :math:`(T \\times 3D)`, where
G
guosheng 已提交
646
              :math:`D` is the hidden size.
647
            - All elements in the weight matrix can be divided into two parts.
G
guosheng 已提交
648
              The first part are weights of the update gate and reset gate with
649
              shape :math:`(D \\times 2D)`, and the second part are weights for
G
guosheng 已提交
650
              candidate hidden state with shape :math:`(D \\times D)`.
651
        bias_attr(ParamAttr): The parameter attribute for learnable the
G
guosheng 已提交
652
            hidden-hidden bias.
653
        is_reverse(bool): Whether to compute reversed GRU, default
G
guosheng 已提交
654 655 656
            :attr:`False`.
        gate_activation(str): The activation for update gate and reset gate.
            Choices = ["sigmoid", "tanh", "relu", "identity"], default "sigmoid".
657
        activation(str): The activation for candidate hidden state.
G
guosheng 已提交
658 659 660
            Choices = ["sigmoid", "tanh", "relu", "identity"], default "tanh".

    Returns:
G
guosheng 已提交
661 662
        Variable: The hidden state of GRU. The shape is :math:`(T \\times D)`, \
            and lod is the same with the input.
663

G
guosheng 已提交
664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706
    Examples:
        .. code-block:: python

            hidden_dim = 512
            x = fluid.layers.fc(input=data, size=hidden_dim * 3)
            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)
    inputs = {'Input': input, 'Weight': weight, 'Bias': bias}
    if h_0 != None:
        assert h_0.shape == (
            size, size), 'The shape of h0 should be(%d, %d)' % (size, size)
        inputs['h0'] = h_0

    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
707 708 709
def gru_unit(input,
             hidden,
             size,
710 711
             param_attr=None,
             bias_attr=None,
Y
Yu Yang 已提交
712
             activation='tanh',
713
             gate_activation='sigmoid'):
Y
Yu Yang 已提交
714
    """
715
    GRU unit layer. The equation of a gru step is:
Y
Yu Yang 已提交
716

717 718
        .. math::
            u_t & = actGate(xu_{t} + W_u h_{t-1} + b_u)
Y
Yu Yang 已提交
719

720
            r_t & = actGate(xr_{t} + W_r h_{t-1} + b_r)
Y
Yu Yang 已提交
721

722
            m_t & = actNode(xm_t + W_c dot(r_t, h_{t-1}) + b_m)
723

724
            h_t & = dot((1-u_t), m_t) + dot(u_t, h_{t-1})
725 726

    The inputs of gru unit includes :math:`z_t`, :math:`h_{t-1}`. In terms
727 728 729
    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
730 731
    connected layer has to be applied, such that :math:`z_t = W_{fc}x_t`.

732 733
    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
734 735 736
    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`.
737 738 739 740 741

    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.
742 743
        param_attr (ParamAttr): The weight parameters for gru unit. Default: None
        bias_attr (ParamAttr): The bias parameters for gru unit. Default: None
744 745 746 747
        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
748

749 750 751 752 753 754
    Returns:
        tuple: The hidden value, reset-hidden value and gate values.

    Examples:

        .. code-block:: python
Y
Yu Yang 已提交
755

756
             # assuming we have x_t_data and prev_hidden of size=10
757
             x_t = fluid.layers.fc(input=x_t_data, size=30)
758 759
             hidden_val, r_h_val, gate_val = fluid.layers.gru_unit(input=x_t,
                                                    hidden = prev_hidden)
Y
Yu Yang 已提交
760 761 762 763 764 765 766 767 768 769 770 771 772 773 774

    """
    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
775 776
    weight = helper.create_parameter(
        attr=helper.param_attr, shape=[size, 3 * size], dtype=dtype)
Y
Yu Yang 已提交
777

778 779 780 781
    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
782
    # create bias
783
    if helper.bias_attr:
Y
Yu Yang 已提交
784 785 786
        bias_size = [1, 3 * size]
        bias = helper.create_parameter(
            attr=helper.bias_attr, shape=bias_size, dtype=dtype, is_bias=True)
787
        inputs['Bias'] = bias
Y
Yu Yang 已提交
788 789 790

    helper.append_op(
        type='gru_unit',
791
        inputs=inputs,
Y
Yu Yang 已提交
792 793 794 795 796 797
        outputs={
            'Gate': gate,
            'ResetHiddenPrev': reset_hidden_pre,
            'Hidden': updated_hidden,
        },
        attrs={
798 799
            'activation': 2,  # tanh
            'gate_activation': 1,  # sigmoid
Y
Yu Yang 已提交
800 801 802 803 804
        })

    return updated_hidden, reset_hidden_pre, gate


Y
yuyang18 已提交
805
@templatedoc()
806
def linear_chain_crf(input, label, param_attr=None):
Y
yuyang18 已提交
807 808 809 810 811 812 813 814 815 816 817 818 819 820
    """
    Linear Chain CRF.

    ${comment}

    Args:
        input(${emission_type}): ${emission_comment}
        label(${label_type}): ${label_comment}
        param_attr(ParamAttr): The attribute of the learnable parameter.

    Returns:
        ${log_likelihood_comment}

    """
Y
Yu Yang 已提交
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
    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
846
@templatedoc()
847
def crf_decoding(input, param_attr, label=None):
Y
yuyang18 已提交
848 849 850 851 852 853 854 855 856 857 858
    """
    ${comment}

    Args:
        input(${emission_type}): ${emission_comment}
        param_attr(ParamAttr): The parameter attribute for training.
        label(${label_type}): ${label_comment}

    Returns:
        ${viterbi_path_comment}
    """
Y
Yu Yang 已提交
859 860 861 862 863 864 865 866 867 868 869 870 871
    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


F
fengjiayi 已提交
872
def cos_sim(X, Y):
Y
Yu Yang 已提交
873 874 875 876
    """
    This function performs the cosine similarity between two tensors
    X and Y and returns that as the output.
    """
F
fengjiayi 已提交
877
    helper = LayerHelper('cos_sim', **locals())
Y
Yu Yang 已提交
878 879 880 881 882 883 884 885 886 887 888 889 890
    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


891
def dropout(x, dropout_prob, is_test=False, seed=None, name=None):
892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908
    """
    Computes dropout.

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

    Args:
       x(variable): The input tensor.
       dropout_prob(float): Probability of setting units to zero.
       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.
909 910
       name(str|None): A name for this layer(optional). If set None, the layer
                    will be named automatically.
911 912 913 914 915 916 917 918 919 920 921

    Returns:
        Variable: A tensor variable.

    Examples:
        .. code-block:: python

          x = fluid.layers.data(name="data", shape=[32, 32], dtype="float32")
          droped = fluid.layers.dropout(input=x, dropout_rate=0.5)
    """

F
fengjiayi 已提交
922
    helper = LayerHelper('dropout', **locals())
923 924 925 926 927 928 929
    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]},
930 931 932 933 934 935
        attrs={
            'dropout_prob': dropout_prob,
            'is_test': is_test,
            'fix_seed': seed is not None,
            'seed': seed if seed is not None else 0
        })
936 937 938
    return out


F
fengjiayi 已提交
939
def cross_entropy(input, label, soft_label=False):
Y
Yu Yang 已提交
940
    """
Y
Yibing Liu 已提交
941 942
    **Cross Entropy Layer**

943 944 945
    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
946 947

    1) One-hot cross-entropy:
F
fengjiayi 已提交
948
        `soft_label = False`, `Label[i, 0]` indicates the class index for sample i:
Y
yangyaming 已提交
949

Y
Yibing Liu 已提交
950
        .. math::
Y
yangyaming 已提交
951

Y
Yibing Liu 已提交
952 953 954
            Y[i] = -\log(X[i, Label[i]])

    2) Soft-label cross-entropy:
F
fengjiayi 已提交
955 956
        `soft_label = True`, `Label[i, j]` indicates the soft label of class j
        for sample i:
Y
Yibing Liu 已提交
957 958 959 960 961

        .. math::

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

Y
Yibing Liu 已提交
962
       Please make sure that in this case the summation of each row of `label`
Y
Yibing Liu 已提交
963 964 965
       equals one.

    3) One-hot cross-entropy with vecterized `label`:
F
fengjiayi 已提交
966 967
         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
968
         to a one-hot cross-entropy with one-hot label representation.
Y
yangyaming 已提交
969

Y
Yibing Liu 已提交
970
    Args:
Y
yangyaming 已提交
971
        input (Variable|list):  a 2-D tensor with shape [N x D], where N is the
972 973 974 975
                                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
976
        label (Variable|list): the ground truth which is a 2-D tensor. When
977 978 979 980
                               `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
981
        soft_label (bool): a flag indicating whether to
982 983
                                           interpretate the given labels as soft
                                           labels, default `False`.
Y
Yibing Liu 已提交
984 985 986 987 988

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

    Raises:
989 990 991 992 993
        `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
994 995 996 997 998 999

    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
1000
    """
F
fengjiayi 已提交
1001
    helper = LayerHelper('cross_entropy', **locals())
Y
Yu Yang 已提交
1002 1003 1004 1005 1006 1007
    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
1008
        attrs={"soft_label": soft_label})
Y
Yu Yang 已提交
1009 1010 1011
    return out


F
fengjiayi 已提交
1012
def square_error_cost(input, label):
Y
Yu Yang 已提交
1013
    """
1014 1015
    **Square error cost layer**

1016 1017
    This layer accepts input predictions and target label and returns the
    squared error cost.
Y
ying 已提交
1018

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
    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:
       input(Variable): Input tensor, has predictions.
       label(Variable): Label tensor, has target labels.

    Returns:
G
guosheng 已提交
1036
        Variable: The tensor variable storing the element-wise squared error \
1037
                  difference of input and label.
1038 1039 1040 1041 1042 1043 1044 1045

    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
1046
    """
F
fengjiayi 已提交
1047
    helper = LayerHelper('square_error_cost', **locals())
Y
Yu Yang 已提交
1048 1049 1050 1051 1052 1053 1054 1055 1056
    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
1057 1058
        type='square', inputs={'X': [minus_out]},
        outputs={'Out': [square_out]})
Y
Yu Yang 已提交
1059 1060 1061 1062 1063 1064 1065
    return square_out


def chunk_eval(input,
               label,
               chunk_scheme,
               num_chunk_types,
F
fengjiayi 已提交
1066
               excluded_chunk_types=None):
Y
Yu Yang 已提交
1067
    """
Y
yangyaming 已提交
1068
    This function computes and outputs the precision, recall and
1069
    F1-score of chunk detection.
Y
Yu Yang 已提交
1070
    """
F
fengjiayi 已提交
1071
    helper = LayerHelper("chunk_eval", **locals())
Y
Yu Yang 已提交
1072 1073 1074 1075 1076

    # prepare output
    precision = helper.create_tmp_variable(dtype="float32")
    recall = helper.create_tmp_variable(dtype="float32")
    f1_score = helper.create_tmp_variable(dtype="float32")
1077 1078 1079
    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
1080 1081 1082 1083 1084 1085 1086 1087

    helper.append_op(
        type="chunk_eval",
        inputs={"Inference": [input],
                "Label": [label]},
        outputs={
            "Precision": [precision],
            "Recall": [recall],
1088 1089 1090 1091
            "F1-Score": [f1_score],
            "NumInferChunks": [num_infer_chunks],
            "NumLabelChunks": [num_label_chunks],
            "NumCorrectChunks": [num_correct_chunks]
Y
Yu Yang 已提交
1092 1093 1094
        },
        attrs={
            "num_chunk_types": num_chunk_types,
G
guosheng 已提交
1095 1096
            "chunk_scheme": chunk_scheme,
            "excluded_chunk_types": excluded_chunk_types or []
Y
Yu Yang 已提交
1097
        })
1098 1099
    return (precision, recall, f1_score, num_infer_chunks, num_label_chunks,
            num_correct_chunks)
Y
Yu Yang 已提交
1100 1101 1102 1103 1104 1105 1106 1107 1108


def sequence_conv(input,
                  num_filters,
                  filter_size=3,
                  filter_stride=1,
                  padding=None,
                  bias_attr=None,
                  param_attr=None,
1109
                  act=None):
Y
Yu Yang 已提交
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
    """
    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.
    """

    # FIXME(dzh) : want to unify the argument of python layer
    # function. So we ignore some unecessary attributes.
    # such as, padding_trainable, context_start.

    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)


1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
def sequence_softmax(input, param_attr=None, bias_attr=None, use_cudnn=True):
    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


1155
def softmax(input, param_attr=None, bias_attr=None, use_cudnn=True, name=None):
1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
    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
1167 1168 1169
def conv2d(input,
           num_filters,
           filter_size,
C
chengduoZH 已提交
1170 1171
           stride=1,
           padding=0,
1172
           dilation=1,
Y
Yu Yang 已提交
1173 1174 1175
           groups=None,
           param_attr=None,
           bias_attr=None,
C
chengduoZH 已提交
1176
           use_cudnn=True,
1177
           use_mkldnn=False,
1178 1179
           act=None,
           name=None):
Y
Yu Yang 已提交
1180
    """
C
chengduoZH 已提交
1181 1182 1183
    **Convlution2D Layer**

    The convolution2D layer calculates the output based on the input, filter
1184 1185 1186
    and strides, paddings, dilations, groups parameters. 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.
C
chengduoZH 已提交
1187 1188
    The details of convolution layer, please refer UFLDL's `convolution,
    <http://ufldl.stanford.edu/tutorial/supervised/FeatureExtractionUsingConvolution/>`_ .
1189 1190 1191
    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
1192

1193
    For each input :math:`X`, the equation is:
C
refine  
chengduoZH 已提交
1194

C
chengduoZH 已提交
1195 1196
    .. math::

C
refine  
chengduoZH 已提交
1197
        Out = \sigma (W \\ast X + b)
C
chengduoZH 已提交
1198

C
chengduoZH 已提交
1199
    In the above equation:
C
chengduoZH 已提交
1200

1201 1202 1203 1204 1205
    * :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.
1206 1207
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be
                   different.
C
chengduoZH 已提交
1208 1209 1210

    Example:

1211 1212
        - Input:

W
weixing02 已提交
1213
          Input shape: :math:`(N, C_{in}, H_{in}, W_{in})`
C
refine  
chengduoZH 已提交
1214

W
weixing02 已提交
1215
          Filter shape: :math:`(C_{out}, C_{in}, H_f, W_f)`
C
refine  
chengduoZH 已提交
1216

1217
        - Output:
W
weixing02 已提交
1218
          Output shape: :math:`(N, C_{out}, H_{out}, W_{out})`
C
refine  
chengduoZH 已提交
1219

C
chengduoZH 已提交
1220
        Where
1221 1222

        .. math::
C
chengduoZH 已提交
1223

W
weixing02 已提交
1224 1225
            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
1226 1227

    Args:
1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
       input(Variable): The input image with [N, C, 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,
           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.
1240 1241 1242
       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.
1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
       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
       act(str): Activation type. Default: None
1253 1254
       name(str|None): A name for this layer(optional). If set None, the layer
           will be named automatically.
C
chengduoZH 已提交
1255 1256

    Returns:
G
guosheng 已提交
1257
        Variable: The tensor variable storing the convolution and \
C
chengduoZH 已提交
1258 1259
                  non-linearity activation result.

C
refine  
chengduoZH 已提交
1260
    Raises:
1261 1262
        ValueError: If the shapes of input, filter_size, stride, padding and
                    groups mismatch.
C
refine  
chengduoZH 已提交
1263

C
chengduoZH 已提交
1264 1265 1266
    Examples:
        .. code-block:: python

1267 1268 1269 1270
          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
1271 1272 1273 1274 1275
    """
    if stride is None:
        stride = [1, 1]

    num_channels = input.shape[1]
1276 1277

    l_type = 'conv2d'
X
xzl 已提交
1278 1279
    if (num_channels == groups and num_filters % num_channels == 0 and
            not use_cudnn):
1280
        l_type = 'depthwise_conv2d'
1281 1282 1283 1284

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

Y
Yu Yang 已提交
1285 1286 1287 1288 1289 1290 1291
    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
1292 1293 1294
    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')
1295
    dilation = utils.convert_to_list(dilation, 2, 'dilation')
C
chengduoZH 已提交
1296

C
chengduoZH 已提交
1297 1298
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
Y
Yu Yang 已提交
1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315

    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(
1316
        type=l_type,
Y
Yu Yang 已提交
1317 1318 1319 1320 1321
        inputs={
            'Input': input,
            'Filter': filter_param,
        },
        outputs={"Output": pre_bias},
C
chengduoZH 已提交
1322 1323 1324
        attrs={
            'strides': stride,
            'paddings': padding,
1325
            'dilations': dilation,
C
chengduoZH 已提交
1326
            'groups': groups,
1327 1328
            'use_cudnn': use_cudnn,
            'use_mkldnn': use_mkldnn
C
chengduoZH 已提交
1329
        })
Y
Yu Yang 已提交
1330 1331 1332 1333 1334 1335

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

    return helper.append_activation(pre_act)


F
fengjiayi 已提交
1336
def sequence_pool(input, pool_type):
Y
Yu Yang 已提交
1337
    """
Y
yangyaming 已提交
1338 1339 1340
    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
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365

    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:
         x.lod = [[0, 2, 5, 7]]
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
         with condition len(x.lod[-1]) - 1 == out.dims[0]

       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)
1366 1367
         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
1368

L
Luo Tao 已提交
1369 1370
    Args:
        input(variable): The input variable which is a LoDTensor.
Y
yangyaming 已提交
1371
        pool_type (string): The pooling type of sequence_pool.
L
Luo Tao 已提交
1372 1373 1374 1375 1376 1377 1378 1379
            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
1380

Y
yangyaming 已提交
1381
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1382 1383 1384 1385 1386
                              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')
1387 1388
             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
1389
    """
F
fengjiayi 已提交
1390
    helper = LayerHelper('sequence_pool', **locals())
Y
Yu Yang 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
    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
1402 1403 1404 1405 1406
    # 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
1407 1408 1409
    return pool_out


F
fengjiayi 已提交
1410
def sequence_first_step(input):
L
Luo Tao 已提交
1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
    """
    This funciton get the first step of sequence.

    .. code-block:: text

       x is a 1-level LoDTensor:
         x.lod = [[0, 2, 5, 7]]
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
         with condition len(x.lod[-1]) - 1 == out.dims[0]
         out.data = [1, 2, 5], where 1=first(1,3), 2=first(2,4,6), 5=first(5,1)
F
fengjiayi 已提交
1425

L
Luo Tao 已提交
1426 1427 1428 1429 1430 1431 1432 1433 1434
    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
1435

Y
yangyaming 已提交
1436
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1437 1438 1439
                              dtype='float32', lod_level=1)
             x_first_step = fluid.layers.sequence_first_step(input=x)
    """
1440 1441 1442
    return sequence_pool(input=input, pool_type="first")


F
fengjiayi 已提交
1443
def sequence_last_step(input):
L
Luo Tao 已提交
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
    """
    This funciton get the last step of sequence.

    .. code-block:: text

       x is a 1-level LoDTensor:
         x.lod = [[0, 2, 5, 7]]
         x.data = [1, 3, 2, 4, 6, 5, 1]
         x.dims = [7, 1]

       then output is a Tensor:
         out.dim = [3, 1]
         with condition len(x.lod[-1]) - 1 == out.dims[0]
         out.data = [3, 6, 1], where 3=last(1,3), 6=last(2,4,6), 1=last(5,1)
F
fengjiayi 已提交
1458

L
Luo Tao 已提交
1459 1460 1461 1462 1463 1464 1465 1466 1467
    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
1468

Y
yangyaming 已提交
1469
             x = fluid.layers.data(name='x', shape=[7, 1],
L
Luo Tao 已提交
1470 1471 1472
                              dtype='float32', lod_level=1)
             x_last_step = fluid.layers.sequence_last_step(input=x)
    """
1473 1474 1475
    return sequence_pool(input=input, pool_type="last")


Y
Yu Yang 已提交
1476
def pool2d(input,
C
chengduoZH 已提交
1477 1478
           pool_size=-1,
           pool_type="max",
C
chengduoZH 已提交
1479 1480
           pool_stride=1,
           pool_padding=0,
C
caoying03 已提交
1481
           global_pooling=False,
C
chengduoZH 已提交
1482
           use_cudnn=True,
1483
           ceil_mode=False,
1484
           use_mkldnn=False,
C
caoying03 已提交
1485
           name=None):
Y
Yu Yang 已提交
1486 1487 1488 1489 1490 1491 1492 1493
    """
    This function adds the operator for pooling in 2 dimensions, using the
    pooling configurations mentioned in input parameters.
    """
    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
1494

C
chengduoZH 已提交
1495 1496 1497 1498 1499
    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
1500 1501 1502 1503
    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
1504 1505
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")
Y
Yu Yang 已提交
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519

    helper = LayerHelper('pool2d', **locals())
    dtype = helper.input_dtype()
    pool_out = helper.create_tmp_variable(dtype)

    helper.append_op(
        type="pool2d",
        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
1520
            "paddings": pool_padding,
1521
            "use_cudnn": use_cudnn,
1522 1523
            "ceil_mode": ceil_mode,
            "use_mkldnn": use_mkldnn
Y
Yu Yang 已提交
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
        })

    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
1536
               data_layout='NCHW',
Y
Yang Yang 已提交
1537
               in_place=False,
1538
               use_mkldnn=False,
1539 1540
               name=None,
               moving_mean_name=None,
W
wanghaoshuang 已提交
1541
               moving_variance_name=None,
W
wanghaoshuang 已提交
1542
               do_model_average_for_mean_and_var=False):
Y
Yu Yang 已提交
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
    """
    This function helps create an operator to implement
    the BatchNorm layer using the configurations from the input parameters.
    """
    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(
1569
        attr=helper.bias_attr, shape=param_shape, dtype=dtype, is_bias=True)
Y
Yu Yang 已提交
1570

1571 1572
    mean = helper.create_parameter(
        attr=ParamAttr(
W
wanghaoshuang 已提交
1573 1574 1575
            name=moving_mean_name,
            initializer=Constant(0.0),
            trainable=False,
W
wanghaoshuang 已提交
1576
            do_model_average=do_model_average_for_mean_and_var),
Q
QI JUN 已提交
1577
        shape=param_shape,
1578 1579 1580 1581 1582 1583 1584
        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
1585
            trainable=False,
W
wanghaoshuang 已提交
1586
            do_model_average=do_model_average_for_mean_and_var),
Q
QI JUN 已提交
1587
        shape=param_shape,
1588 1589
        dtype=input.dtype)
    variance.stop_gradient = True
Y
Yu Yang 已提交
1590 1591 1592 1593 1594 1595

    # 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
1596 1597
    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
1598

Y
Yang Yang 已提交
1599
    batch_norm_out = input if in_place else helper.create_tmp_variable(dtype)
Y
Yu Yang 已提交
1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616

    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
        },
1617 1618 1619 1620 1621 1622
        attrs={
            "momentum": momentum,
            "epsilon": epsilon,
            "is_test": is_test,
            "use_mkldnn": use_mkldnn
        })
Y
Yu Yang 已提交
1623 1624 1625 1626

    return helper.append_activation(batch_norm_out)


G
guosheng 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
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):
    """
    **Layer Normalization**

1639
    Assume feature vectors exist on dimensions
G
guosheng 已提交
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659
    :attr:`begin_norm_axis ... rank(input)` and calculate the moment statistics
    along these dimensions for each feature vector :math:`a` with size
    :math:`H`, then normalize each feature vector using the corresponding
    statistics. After that, apply learnable gain and bias on the normalized
    tensor to scale and shift if :attr:`scale` and :attr:`shift` are set.

    Refer to `Layer Normalization <https://arxiv.org/pdf/1607.06450v1.pdf>`_

    The formula is as follows:

    .. math::

        \\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)

    Args:
        input(Variable): The input tensor variable.
1660
        scale(bool): Whether to learn the adaptive gain :math:`g` after
G
guosheng 已提交
1661
            normalization.
1662
        shift(bool): Whether to learn the adaptive bias :math:`b` after
G
guosheng 已提交
1663
            normalization.
1664
        begin_norm_axis(bool): The normalization will be performed along
G
guosheng 已提交
1665
            dimensions from :attr:`begin_norm_axis` to :attr:`rank(input)`.
1666
        epsilon(float): The small value added to the variance to prevent
G
guosheng 已提交
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
            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.

    Returns:
        Variable: A tensor variable with the same shape as the input.

    Examples:
        .. code-block:: python

            data = fluid.layers.data(
              name='data', shape=[3, 32, 32], dtype='float32')
            x = fluid.layers.layer_norm(input=data, begin_norm_axis=1)
    """
    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
1698
    if shift:
G
guosheng 已提交
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722
        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
1723
def beam_search_decode(ids, scores, name=None):
Y
Yu Yang 已提交
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
    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
1744 1745 1746
                     padding=0,
                     stride=1,
                     dilation=1,
1747
                     groups=None,
C
caoying03 已提交
1748
                     param_attr=None,
1749
                     bias_attr=None,
C
chengduoZH 已提交
1750
                     use_cudnn=True,
1751
                     act=None,
C
caoying03 已提交
1752
                     name=None):
Y
Yu Yang 已提交
1753
    """
1754 1755 1756 1757 1758 1759 1760 1761
    **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
1762 1763
    layer, please refer to the following explanation and references
    `therein <http://www.matthewzeiler.com/wp-content/uploads/2017/07/cvpr2010.pdf>`_.
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775

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

    .. math::

        Out = W \\ast X

    In the above equation:

    * :math:`X`: Input value, a tensor with NCHW format.
    * :math:`W`: Filter value, a tensor with MCHW format.
    * :math:`\\ast` : Convolution transpose operation.
1776 1777
    * :math:`Out`: Output value, the shape of :math:`Out` and :math:`X` may be
                   different.
Y
Yu Yang 已提交
1778

1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791
    Example:

        - Input:

          Input shape: $(N, C_{in}, H_{in}, W_{in})$

          Filter shape: $(C_{in}, C_{out}, H_f, W_f)$

        - Output:

          Output shape: $(N, C_{out}, H_{out}, W_{out})$

        Where
Y
Yu Yang 已提交
1792

1793 1794 1795 1796
        .. 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
1797 1798

    Args:
1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
       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.
1818 1819 1820 1821 1822 1823
       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
1824 1825
       param_attr(ParamAttr): The parameters to the Conv2d_transpose Layer.
                              Default: None
1826
       bias_attr(ParamAttr): Bias parameter for the Conv2d layer. Default: None
1827 1828
       use_cudnn(bool): Use cudnn kernel or not, it is valid only when the cudnn
           library is installed. Default: True
1829
       act(str): Activation type. Default: None
1830 1831
       name(str|None): A name for this layer(optional). If set None, the layer
           will be named automatically.
Y
Yu Yang 已提交
1832 1833

    Returns:
1834 1835 1836
       Variable: The tensor variable storing the convolution transpose result.

    Raises:
1837 1838
       ValueError: If the shapes of input, filter_size, stride, padding and
                   groups mismatch.
1839 1840 1841 1842

    Examples:
       .. code-block:: python

1843 1844 1845 1846
          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
1847 1848 1849 1850 1851 1852
    """
    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
1853 1854 1855
    padding = utils.convert_to_list(padding, 2, 'padding')
    stride = utils.convert_to_list(stride, 2, 'stride')
    dilation = utils.convert_to_list(dilation, 2, 'dilation')
C
chengduoZH 已提交
1856

C
chengduoZH 已提交
1857 1858 1859
    if not isinstance(use_cudnn, bool):
        raise ValueError("use_cudnn should be True or False")

Y
Yu Yang 已提交
1860 1861 1862 1863 1864 1865 1866 1867
    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]

        h_in = input.shape[2]
        w_in = input.shape[3]
C
chengduoZH 已提交
1868 1869 1870 1871 1872

        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
1873
        filter_size = [filter_size_h, filter_size_w]
C
chengduoZH 已提交
1874 1875 1876
    else:
        filter_size = utils.convert_to_list(filter_size, 2,
                                            'conv2d_transpose.filter_size')
Y
Yu Yang 已提交
1877

1878 1879
    groups = 1 if groups is None else groups
    filter_shape = [input_channel, num_filters / groups] + filter_size
Y
Yu Yang 已提交
1880 1881 1882
    img_filter = helper.create_parameter(
        dtype=input.dtype, shape=filter_shape, attr=helper.param_attr)

1883
    pre_bias = helper.create_tmp_variable(dtype=input.dtype)
Y
Yu Yang 已提交
1884 1885 1886 1887
    helper.append_op(
        type='conv2d_transpose',
        inputs={'Input': [input],
                'Filter': [img_filter]},
1888
        outputs={'Output': pre_bias},
C
chengduoZH 已提交
1889 1890 1891 1892
        attrs={
            'strides': stride,
            'paddings': padding,
            'dilations': dilation,
1893
            'groups': groups,
C
chengduoZH 已提交
1894 1895
            'use_cudnn': use_cudnn
        })
Y
Yu Yang 已提交
1896

1897 1898
    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
1899
    return out
Y
yangyaming 已提交
1900 1901


Y
yangyaming 已提交
1902
def sequence_expand(x, y, ref_level=-1, name=None):
1903
    """Sequence Expand Layer. This layer will expand the input variable **x**
Y
yangyaming 已提交
1904 1905 1906 1907
    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:
1908 1909 1910 1911 1912

    .. code-block:: text

        * Case 1
            x is a LoDTensor:
Y
yangyaming 已提交
1913 1914
                x.lod  = [[0,   2,        4]]
                x.data = [[a], [b], [c], [d]]
1915 1916 1917 1918 1919 1920
                x.dims = [4, 1]

            y is a LoDTensor:
                y.lod = [[0,    2,    4],
                         [0, 3, 6, 7, 8]]

Y
yangyaming 已提交
1921
            ref_level: 0
1922

Y
yangyaming 已提交
1923 1924 1925
            then output is a 1-level LoDTensor:
                out.lod =  [[0,   2,        4,        6,        8]]
                out.data = [[a], [b], [a], [b], [c], [d], [c], [d]]
1926 1927 1928 1929
                out.dims = [8, 1]

        * Case 2
            x is a Tensor:
Y
yangyaming 已提交
1930
                x.data = [[a], [b], [c]]
1931 1932 1933
                x.dims = [3, 1]

            y is a LoDTensor:
Y
yangyaming 已提交
1934
                y.lod = [[0, 2, 2, 5]]
1935

Y
yangyaming 已提交
1936
            ref_level: -1
1937

Y
yangyaming 已提交
1938 1939 1940
            then output is a Tensor:
                out.data = [[a], [a], [c], [c], [c]]
                out.dims = [5, 1]
1941 1942 1943
    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
1944 1945
        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
1946
        name(str|None): A name for this layer(optional). If set None, the layer
Y
yangyaming 已提交
1947
                        will be named automatically.
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957

    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
1958
            out = layers.sequence_expand(x=x, y=y, ref_level=0)
1959
    """
Y
yangyaming 已提交
1960
    helper = LayerHelper('sequence_expand', input=x, **locals())
1961 1962 1963
    dtype = helper.input_dtype()
    tmp = helper.create_tmp_variable(dtype)
    helper.append_op(
Y
yangyaming 已提交
1964 1965 1966 1967 1968
        type='sequence_expand',
        inputs={'X': x,
                'Y': y},
        outputs={'Out': tmp},
        attrs={'ref_level': ref_level})
1969
    return tmp
1970 1971


Q
Qiao Longfei 已提交
1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
def beam_search(pre_ids, ids, scores, beam_size, end_id, level=0):
    '''
    This function implements the beam search algorithm.
    '''
    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
2004 2005 2006 2007
def lstm_unit(x_t,
              hidden_t_prev,
              cell_t_prev,
              forget_bias=0.0,
Y
yangyaming 已提交
2008
              param_attr=None,
C
caoying03 已提交
2009 2010
              bias_attr=None,
              name=None):
Y
yangyaming 已提交
2011 2012 2013 2014
    """Lstm unit layer. The equation of a lstm step is:

        .. math::

2015
            i_t & = \sigma(W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i)
Y
yangyaming 已提交
2016

2017
            f_t & = \sigma(W_{x_f}x_{t} + W_{h_f}h_{t-1} + b_f)
Y
yangyaming 已提交
2018

2019
            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
2020

2021
            o_t & = \sigma(W_{x_o}x_{t} + W_{h_o}h_{t-1} + b_o)
Y
yangyaming 已提交
2022 2023 2024

            h_t & = o_t tanh(c_t)

2025 2026 2027 2028 2029 2030
    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
2031 2032 2033

        .. math::

2034
            L_{i_t} = W_{x_i}x_{t} + W_{h_i}h_{t-1} + b_i
Y
yangyaming 已提交
2035 2036 2037 2038 2039 2040 2041 2042

    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
2043
    This layer has two outputs including :math:`h_t` and :math:`o_t`.
Y
yangyaming 已提交
2044 2045

    Args:
Y
yangyaming 已提交
2046 2047 2048 2049 2050 2051
        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
2052
        forget_bias (float): The forget bias of lstm unit.
Y
yangyaming 已提交
2053 2054
        param_attr (ParamAttr): The attributes of parameter weights, used to set
            initializer, name etc.
Y
yangyaming 已提交
2055 2056
        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
2057 2058
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
Y
yangyaming 已提交
2059 2060

    Returns:
Y
yangyaming 已提交
2061
        tuple: The hidden value and cell value of lstm unit.
Y
yangyaming 已提交
2062 2063

    Raises:
2064 2065 2066 2067
        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
2068 2069 2070 2071 2072 2073

    Examples:

        .. code-block:: python

             x_t = fluid.layers.fc(input=x_t_data, size=10)
2074
             prev_hidden = fluid.layers.fc(input=prev_hidden_data, size=30)
Y
yangyaming 已提交
2075
             prev_cell = fluid.layers.fc(input=prev_cell_data, size=30)
Y
yangyaming 已提交
2076
             hidden_value, cell_value = fluid.layers.lstm_unit(x_t=x_t,
Y
yangyaming 已提交
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092
                                                    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
2093
        raise ValueError("The 1st dimensions of x_t, hidden_t_prev and "
2094 2095 2096 2097
                         "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
2098 2099
                         "cell_t_prev must be the same.")

Y
yangyaming 已提交
2100 2101 2102
    if bias_attr is None:
        bias_attr = ParamAttr()

Y
yangyaming 已提交
2103
    size = cell_t_prev.shape[1]
2104
    concat_out = concat(input=[x_t, hidden_t_prev], axis=1)
Y
yangyaming 已提交
2105 2106
    fc_out = fc(input=concat_out,
                size=4 * size,
Y
yangyaming 已提交
2107
                param_attr=param_attr,
2108
                bias_attr=bias_attr)
Y
yangyaming 已提交
2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120
    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
2121
    return h, c
G
guosheng 已提交
2122 2123


C
caoying03 已提交
2124
def reduce_sum(input, dim=None, keep_dim=False, name=None):
G
guosheng 已提交
2125
    """
Y
yangyaming 已提交
2126
    Computes the sum of tensor elements over the given dimension.
G
guosheng 已提交
2127 2128 2129

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2130
        dim (list|int|None): The dimensions along which the sum is performed. If
Y
yangyaming 已提交
2131 2132
            :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
2133 2134
            range :math:`[-rank(input), rank(input))`. If :math:`dim[i] < 0`,
            the dimension to reduce is :math:`rank + dim[i]`.
2135
        keep_dim (bool|False): Whether to reserve the reduced dimension in the
Y
yangyaming 已提交
2136
            output Tensor. The result tensor will have one fewer dimension
G
guosheng 已提交
2137
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2138 2139
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
G
guosheng 已提交
2140 2141 2142

    Returns:
        Variable: The reduced Tensor variable.
F
fengjiayi 已提交
2143

G
guosheng 已提交
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154
    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
2155 2156 2157 2158 2159 2160 2161 2162

            # 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
2163 2164 2165
    """
    helper = LayerHelper('reduce_sum', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2166 2167
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
G
guosheng 已提交
2168 2169 2170 2171 2172
    helper.append_op(
        type='reduce_sum',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2173
            'dim': dim if dim != None else [0],
G
guosheng 已提交
2174 2175 2176 2177
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
G
guosheng 已提交
2178 2179


C
caoying03 已提交
2180
def reduce_mean(input, dim=None, keep_dim=False, name=None):
G
guosheng 已提交
2181
    """
Y
yangyaming 已提交
2182
    Computes the mean of tensor elements over the given dimension.
G
guosheng 已提交
2183 2184 2185

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2186
        dim (list|int|None): The dimensions along which the mean is computed. If
Y
yangyaming 已提交
2187 2188 2189
            :attr:`None`, compute the mean 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))`. If
W
whs 已提交
2190
            :math:`dim[i] < 0`, the dimension to reduce is :math:`rank + dim[i]`.
Y
yangyaming 已提交
2191 2192
        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
2193
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2194 2195
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
G
guosheng 已提交
2196 2197 2198

    Returns:
        Variable: The reduced Tensor variable.
F
fengjiayi 已提交
2199

G
guosheng 已提交
2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
    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
2210 2211
            fluid.layers.reduce_mean(
                x, dim=1, keep_dim=True)  # [[0.475], [0.4]]
W
whs 已提交
2212 2213 2214 2215 2216 2217 2218

            # 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
2219 2220 2221
    """
    helper = LayerHelper('reduce_mean', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2222 2223
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
G
guosheng 已提交
2224 2225 2226 2227 2228
    helper.append_op(
        type='reduce_mean',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2229
            'dim': dim if dim != None else [0],
G
guosheng 已提交
2230 2231 2232 2233
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
2234 2235


C
caoying03 已提交
2236
def reduce_max(input, dim=None, keep_dim=False, name=None):
2237
    """
Y
yangyaming 已提交
2238
    Computes the maximum of tensor elements over the given dimension.
2239 2240 2241

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2242
        dim (list|int|None): The dimension along which the maximum is computed.
Y
yangyaming 已提交
2243 2244 2245
            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
2246
            If :math:`dim[i] < 0`, the dimension to reduce is :math:`rank + dim[i]`.
Y
yangyaming 已提交
2247 2248
        keep_dim (bool): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
2249
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2250 2251
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
2252 2253 2254

    Returns:
        Variable: The reduced Tensor variable.
Y
yangyaming 已提交
2255

2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
    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
2267 2268 2269 2270 2271 2272 2273

            # 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]
2274 2275 2276
    """
    helper = LayerHelper('reduce_max', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2277 2278
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
2279 2280 2281 2282 2283
    helper.append_op(
        type='reduce_max',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2284
            'dim': dim if dim != None else [0],
2285 2286 2287 2288 2289 2290
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out


C
caoying03 已提交
2291
def reduce_min(input, dim=None, keep_dim=False, name=None):
2292
    """
Y
yangyaming 已提交
2293
    Computes the minimum of tensor elements over the given dimension.
2294 2295 2296

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
W
whs 已提交
2297
        dim (list|int|None): The dimensions along which the minimum is computed.
Y
yangyaming 已提交
2298 2299 2300
            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
2301
            If :math:`dim[i] < 0`, the dimension to reduce is :math:`rank + dim[i]`.
Y
yangyaming 已提交
2302 2303
        keep_dim (bool): Whether to reserve the reduced dimension in the
            output Tensor. The result tensor will have one fewer dimension
2304
            than the :attr:`input` unless :attr:`keep_dim` is true.
C
caoying03 已提交
2305 2306
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
2307 2308 2309

    Returns:
        Variable: The reduced Tensor variable.
Y
yangyaming 已提交
2310

2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321
    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
2322 2323 2324 2325 2326 2327 2328

            # 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]
2329 2330 2331
    """
    helper = LayerHelper('reduce_min', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2332 2333
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
2334 2335 2336 2337 2338
    helper.append_op(
        type='reduce_min',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2339
            'dim': dim if dim != None else [0],
2340 2341 2342 2343
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out
G
guosheng 已提交
2344 2345


2346 2347 2348 2349 2350 2351
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
2352
        dim (list|int|None): The dimensions along which the product is performed. If
2353 2354
            :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
2355 2356
            range :math:`[-rank(input), rank(input))`. If :math:`dim[i] < 0`,
            the dimension to reduce is :math:`rank + dim[i]`.
2357 2358 2359
        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
2360
        name(str|None): A name for this layer(optional). If set None, the
Z
zhouhanqing 已提交
2361
            layer will be named automatically.
2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375

    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
2376
            fluid.layers.reduce_prod(x, dim=1,
Z
zhouhanqing 已提交
2377
                                     keep_dim=True)  # [[0.027], [0.0084]]
W
whs 已提交
2378 2379 2380 2381 2382 2383 2384

            # 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]
2385 2386 2387
    """
    helper = LayerHelper('reduce_prod', **locals())
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
W
whs 已提交
2388 2389
    if dim is not None and not isinstance(dim, list):
        dim = [dim]
2390 2391 2392 2393 2394
    helper.append_op(
        type='reduce_prod',
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
whs 已提交
2395
            'dim': dim if dim != None else [0],
2396 2397 2398 2399 2400 2401
            'keep_dim': keep_dim,
            'reduce_all': True if dim == None else False
        })
    return out


C
caoying03 已提交
2402
def split(input, num_or_sections, dim=-1, name=None):
G
guosheng 已提交
2403
    """
C
caoying03 已提交
2404
    Split the input tensor into multiple sub-tensors.
G
guosheng 已提交
2405 2406 2407

    Args:
        input (Variable): The input variable which is a Tensor or LoDTensor.
C
caoying03 已提交
2408 2409 2410 2411 2412
        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
2413
            :attr:`dim` dimension orderly.
C
caoying03 已提交
2414
        dim (int): The dimension along which to split. If :math:`dim < 0`, the
G
guosheng 已提交
2415
            dimension to split along is :math:`rank(input) + dim`.
C
caoying03 已提交
2416 2417
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
G
guosheng 已提交
2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429

    Returns:
        List: The list of segmented tensor variables.

    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
2430 2431
            x0, x1, x2 = fluid.layers.split(
                x, num_or_sections=[2, 3, 4], dim=1)
G
guosheng 已提交
2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460
            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
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493


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

    output = x / sqrt(max(sum(x**2), epsilon))

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

    Args:
       x(Variable|list): The input tensor to l2_normalize layer.
       axis(int): Dimension along which to normalize the input.
       epsilon(float): A lower bound value for `x`'s l2 norm. sqrt(epsilon) will
                       be used as the divisor if the l2 norm of `x` is less than
                       sqrt(epsilon).
       name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.


    Returns:
        Variable: The output tensor variable.

    Examples:
        .. code-block:: python

          data = fluid.layers.data(name="data",
                                   shape=(3, 17, 13),
                                   dtype="float32")
Y
ying 已提交
2494
          normed = fluid.layers.l2_normalize(x=data, axis=1)
C
caoying03 已提交
2495 2496
    """

F
fengjiayi 已提交
2497 2498
    if len(x.shape) == 1:
        axis = 0
C
caoying03 已提交
2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509
    helper = LayerHelper("l2_normalize", **locals())

    square = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(type="square", inputs={"X": x}, outputs={"Out": square})

    reduced_sum = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type="reduce_sum",
        inputs={"X": square},
        outputs={"Out": reduced_sum},
        attrs={
W
whs 已提交
2510
            "dim": [1] if axis is None else [axis],
C
caoying03 已提交
2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
            "keep_dim": True,
            "reduce_all": False
        })

    # TODO(caoying) A lower bound value epsilon for the norm is needed to
    # imporve the numeric stability of reciprocal. This requires a maximum_op.
    rsquare = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type="reciprocal", inputs={"X": reduced_sum}, outputs={"Out": rsquare})

    # TODO(caoying) the current elementwise_mul operator does not support a
    # general broadcast rule which broadcasts input(Y) to have the same
    # dimension with Input(X) starting from a specified dimension. So this
2524
    # exanpsion is requred. Once a general broadcast rule is spported, this
C
caoying03 已提交
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541
    # expanding canbe removed.
    rsquare_expanded = helper.create_tmp_variable(dtype=x.dtype)
    expand_times = [1] * len(x.shape)
    expand_times[axis] = int(x.shape[axis])
    helper.append_op(
        type="expand",
        inputs={"X": rsquare},
        outputs={"Out": rsquare_expanded},
        attrs={"expand_times": expand_times})

    out = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type="elementwise_mul",
        inputs={"X": x,
                "Y": rsquare_expanded},
        outputs={"Out": out})
    return out
2542 2543


2544
def matmul(x, y, transpose_x=False, transpose_y=False, name=None):
G
guosheng 已提交
2545
    """
Y
ying 已提交
2546 2547 2548 2549
    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
2550

C
chengduoZH 已提交
2551
    The actual behavior depends on the shapes of :math:`x`, :math:`y` and the
2552
    flag values of :attr:`transpose_x`, :attr:`transpose_y`. Specifically:
G
guosheng 已提交
2553

2554 2555 2556 2557 2558
    - 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
2559
      :math:`[1, D]` in transposed form.
G
guosheng 已提交
2560

C
chengduoZH 已提交
2561
    - After transpose, the two tensors are 2-D or n-D and matrix multiplication
2562
      performs in the following way.
G
guosheng 已提交
2563

2564
      - If both are 2-D, they are multiplied like conventional matrices.
C
chengduoZH 已提交
2565
      - If either is n-D, it is treated as a stack of matrices residing in the
Y
ying 已提交
2566
        last two dimensions and a batched matrix multiply supporting broadcast
2567
        applies on the two tensors.
G
guosheng 已提交
2568

Y
ying 已提交
2569 2570
    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
2571
    removed after matrix multiplication.
G
guosheng 已提交
2572 2573 2574

    Args:
        x (Variable): The input variable which is a Tensor or LoDTensor.
2575 2576 2577
        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.
2578
        name(str|None): A name for this layer(optional). If set None, the layer
2579
            will be named automatically.
G
guosheng 已提交
2580 2581

    Returns:
2582
        Variable: The product Tensor variable.
G
guosheng 已提交
2583

G
guosheng 已提交
2584 2585 2586
    Examples:
        .. code-block:: python

2587
            # Examples to clarify shapes of the inputs and output
C
chengduoZH 已提交
2588 2589
            # x: [B, ..., M, K], y: [B, ..., K, N]
            fluid.layers.matmul(x, y)  # out: [B, ..., M, N]
Y
ying 已提交
2590

2591 2592
            # x: [B, M, K], y: [B, K, N]
            fluid.layers.matmul(x, y)  # out: [B, M, N]
Y
ying 已提交
2593

2594 2595
            # x: [B, M, K], y: [K, N]
            fluid.layers.matmul(x, y)  # out: [B, M, N]
Y
ying 已提交
2596

2597 2598
            # x: [M, K], y: [K, N]
            fluid.layers.matmul(x, y)  # out: [M, N]
Y
ying 已提交
2599 2600 2601 2602

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

2603 2604
            # x: [K], y: [K]
            fluid.layers.matmul(x, y)  # out: [1]
2605

Y
ying 已提交
2606
            # x: [M], y: [N]
2607
            fluid.layers.matmul(x, y, True, True)  # out: [M, N]
G
guosheng 已提交
2608
    """
Y
ying 已提交
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620

    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
2621
            y_shape = y_shape + [1]
Y
ying 已提交
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637

        # 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)

2638
    helper = LayerHelper('matmul', **locals())
Y
ying 已提交
2639
    out = helper.create_tmp_variable(dtype=x.dtype)
G
guosheng 已提交
2640
    helper.append_op(
2641 2642 2643 2644 2645 2646 2647
        type='matmul',
        inputs={'X': x,
                'Y': y},
        outputs={'Out': out},
        attrs={'transpose_X': transpose_x,
               'transpose_Y': transpose_y})
    return out
2648 2649


2650
def topk(input, k, name=None):
Q
qingqing01 已提交
2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665
    """
    This operator is used to find values and indices of the k largest entries
    for the last dimension.

    If the input is a vector (rank=1), finds the k largest entries in the vector
    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.

    Args:
        input(Variable): The input variable which can be a vector or Tensor with
            higher rank.
        k(int): An integer value to specify the top k largest elements.
2666 2667
        name(str|None): A name for this layer(optional). If set None, the layer
                       will be named automatically.
Q
qingqing01 已提交
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698

    Returns:
        values(Variable): The k largest elements along each last dimensional
            slice.
        indices(Variable): The indices of values within the last dimension of
            input.

    Examples:
        .. code-block:: python

            top5_values, top5_indices = layers.topk(input, k=5)
    """
    shape = input.shape
    if k < 1 and k >= shape[-1]:
        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


W
wanghaoshuang 已提交
2699
def edit_distance(input, label, normalized=True, ignored_tokens=None,
W
wanghaoshuang 已提交
2700
                  name=None):
2701
    """
Y
ying 已提交
2702 2703 2704 2705 2706 2707 2708 2709 2710
    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
2711

Y
ying 已提交
2712
    "kitten" -> "sitten" -> "sittin" -> "sitting"
W
wanghaoshuang 已提交
2713

Y
ying 已提交
2714 2715 2716 2717
    Input(Hyps) is a LoDTensor consisting of all the hypothesis strings with
    the total number denoted by `batch_size`, and the separation is specified
    by the LoD information. And the `batch_size` reference strings are arranged
    in order in the same way in the LoDTensor Input(Refs).
W
wanghaoshuang 已提交
2718

Y
ying 已提交
2719 2720 2721
    Output(Out) contains the `batch_size` results and each stands for the edit
    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
2722

2723 2724 2725 2726 2727
    Args:

        input(Variable): The indices for hypothesis strings.

        label(Variable): The indices for reference strings.
W
wanghaoshuang 已提交
2728

Y
ying 已提交
2729 2730
        normalized(bool): Indicated whether to normalize the edit distance by
                          the length of reference string.
2731

Y
ying 已提交
2732 2733
        ignored_tokens(list of int): Tokens that should be removed before
                                     calculating edit distance.
2734

W
wanghaoshuang 已提交
2735
    Returns:
W
wanghaoshuang 已提交
2736
        Variable: sequence-to-sequence edit distance in shape [batch_size, 1].
W
wanghaoshuang 已提交
2737 2738 2739 2740 2741

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[8], dtype='float32')
2742 2743
            y = fluid.layers.data(name='y', shape=[7], dtype='float32')

2744
            cost = fluid.layers.edit_distance(input=x,label=y)
2745
    """
2746
    helper = LayerHelper("edit_distance", **locals())
2747

2748
    # remove some tokens from input and labels
W
wanghaoshuang 已提交
2749
    if ignored_tokens is not None and len(ignored_tokens) > 0:
2750 2751 2752 2753 2754 2755 2756
        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
2757
            attrs={"tokens": ignored_tokens})
2758 2759 2760 2761 2762
        input = erased_input

        helper.append_op(
            type="sequence_erase",
            inputs={"X": [label]},
W
whs 已提交
2763
            outputs={"Out": [erased_label]},
W
wanghaoshuang 已提交
2764
            attrs={"tokens": ignored_tokens})
2765 2766
        label = erased_label

2767 2768
    # edit distance op
    edit_distance_out = helper.create_tmp_variable(dtype="int64")
2769
    sequence_num = helper.create_tmp_variable(dtype="int64")
2770 2771 2772 2773
    helper.append_op(
        type="edit_distance",
        inputs={"Hyps": [input],
                "Refs": [label]},
2774 2775
        outputs={"Out": [edit_distance_out],
                 "SequenceNum": [sequence_num]},
2776 2777
        attrs={"normalized": normalized})

2778
    return edit_distance_out, sequence_num
2779 2780 2781 2782 2783


def ctc_greedy_decoder(input, blank, name=None):
    """
    This op is used to decode sequences by greedy policy by below steps:
Y
ying 已提交
2784 2785 2786 2787
    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.
2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816

    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]]

        input.lod = [[0, 4, 8]]

        Then:

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

        output.lod = [[0, 2, 3]]

    Args:

Y
ying 已提交
2817 2818 2819 2820 2821 2822
        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).
2823

Y
ying 已提交
2824 2825 2826
        blank(int): the blank label index of Connectionist Temporal
                    Classification (CTC) loss, which is in thehalf-opened
                    interval [0, num_classes + 1).
2827 2828

    Returns:
2829
        Variable: CTC greedy decode result. If all the sequences in result were
2830
        empty, the result LoDTensor will be [-1] with LoD [[0]] and dims [1, 1].
2831 2832 2833 2834 2835

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[8], dtype='float32')
W
wanghaoshuang 已提交
2836

2837
            cost = fluid.layers.ctc_greedy_decoder(input=x, blank=0)
W
wanghaoshuang 已提交
2838
    """
2839
    helper = LayerHelper("ctc_greedy_decoder", **locals())
Q
qingqing01 已提交
2840
    _, topk_indices = topk(input, k=1)
2841 2842 2843 2844 2845 2846

    # 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
2847
        outputs={"Output": [ctc_out]},
2848 2849
        attrs={"merge_repeated": True,
               "blank": blank})
2850
    return ctc_out
2851 2852


F
fengjiayi 已提交
2853
def warpctc(input, label, blank=0, norm_by_times=False):
W
wanghaoshuang 已提交
2854
    """
2855 2856
    An operator integrating the open source Warp-CTC library
    (https://github.com/baidu-research/warp-ctc)
W
wanghaoshuang 已提交
2857
    to compute Connectionist Temporal Classification (CTC) loss.
2858 2859
    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
2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872
    input tensor.

    Args:
       input(Variable): (LodTensor, default: LoDTensor<float>),
         the unscaled 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).
       label(Variable): (LodTensor, default: LoDTensor<int>), 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.
2873
       blank: (int, default: 0), the blank label index of Connectionist
W
wanghaoshuang 已提交
2874 2875
         Temporal Classification (CTC) loss, which is in the
         half-opened interval [0, num_classes + 1).
2876
       norm_by_times: (bool, default: false), whether to normalize
W
wanghaoshuang 已提交
2877
       the gradients by the number of time-step, which is also the
2878 2879
       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
2880 2881

    Returns:
2882 2883
        Variable: The Connectionist Temporal Classification (CTC) loss,
        which is a 2-D Tensor of the shape [batch_size, 1].
W
wanghaoshuang 已提交
2884 2885 2886

    Examples:
        .. code-block:: python
2887 2888 2889 2890
            y = layers.data(
                name='y', shape=[11, 8], dtype='float32', lod_level=1)
            y_predict = layers.data(
                name='y_predict', shape=[11, 1], dtype='float32')
W
wanghaoshuang 已提交
2891 2892 2893
            cost = layers.warpctc(input=y_predict, label=y)

    """
F
fengjiayi 已提交
2894
    helper = LayerHelper('warpctc', **locals())
W
wanghaoshuang 已提交
2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905
    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
2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959


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]]
            x.data = [[1, 2], [3, 4],
                      [5, 6], [7, 8], [9, 10], [11, 12]]
            x.dims = [6, 2]

        set new_dim = 4

        then out is a LoDTensor:
            out.lod  = [[0, 1, 3]]
            out.data = [[1, 2, 3, 4],
                        [5, 6, 7, 8], [9, 10, 11, 12]]
            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:
       input (Variable): (LodTensor, default: LoDTensor<float>), a 2-D LoDTensor
                with shape being [N, M] where M for dimension.
       new_dim (int): New dimension which the input LoDTensor is reshaped to.

    Returns:
        Variable: Reshaped LoDTensor according to new dimension.

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[5, 20],
                              dtype='float32', lod_level=1)
            x_reshaped = layers.sequence_reshape(input=x, new_dim=10)
    """
    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
2960 2961


2962
@autodoc()
Y
Yang Yu 已提交
2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988
def nce(input,
        label,
        num_total_classes,
        sample_weight=None,
        param_attr=None,
        bias_attr=None,
        num_neg_samples=None):
    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
2989 2990 2991 2992 2993 2994 2995 2996 2997
    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
2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013

    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
3014
    return cost / (num_neg_samples + 1)
3015 3016


Y
fix ci.  
ying 已提交
3017
def transpose(x, perm, name=None):
Y
ying 已提交
3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036
    """
    **transpose Layer**

    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:
       input (Variable): (Tensor), A Tensor.
       perm (list): A permutation of the dimensions of `input`.

    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
3037
            x_transposed = layers.transpose(x, perm=[1, 0, 2])
Y
ying 已提交
3038 3039
    """

Y
fix ci.  
ying 已提交
3040
    if len(perm) != len(x.shape):
Y
ying 已提交
3041 3042 3043
        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
3044 3045 3046 3047 3048 3049
    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
3050 3051

    helper = LayerHelper('transpose', **locals())
Y
fix ci.  
ying 已提交
3052
    out = helper.create_tmp_variable(x.dtype)
Y
ying 已提交
3053 3054
    helper.append_op(
        type='transpose',
Y
fix ci.  
ying 已提交
3055
        inputs={'X': [x]},
Y
ying 已提交
3056 3057 3058
        outputs={'Out': [out]},
        attrs={'axis': perm})
    return out
3059 3060


3061
def im2sequence(input, filter_size=1, stride=1, padding=0, name=None):
3062
    """
3063 3064 3065 3066 3067 3068 3069
    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:
3070 3071 3072 3073 3074 3075 3076 3077 3078 3079

    .. 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
3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097

        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.

3098 3099 3100
        name (int): The name of this layer. It is optional.

    Returns:
W
wanghaoshuang 已提交
3101 3102 3103 3104 3105
        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.
3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134

    Examples:

    As an example:

        .. 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
3135 3136 3137
            filter = [2, 2]
            stride = [1, 1]
            padding = [0, 0]
3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157

            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}

            output.lod = [[0, 4, 8]]

        The simple usage is:

        .. code-block:: python

3158 3159
            output = fluid.layers.im2sequence(
                input=layer, stride=[1, 1], filter_size=[2, 2])
3160 3161

    """
W
wanghaoshuang 已提交
3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172

    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])

3173
    helper = LayerHelper('im2sequence', **locals())
3174 3175
    out = helper.create_tmp_variable(dtype=helper.input_dtype())
    helper.append_op(
3176
        type='im2sequence',
3177 3178 3179
        inputs={'X': input},
        outputs={'Out': out},
        attrs={
W
wanghaoshuang 已提交
3180 3181 3182
            'kernels': filter_size,
            'strides': stride,
            'paddings': padding,
3183 3184
        })
    return out
3185 3186


3187 3188 3189 3190
def row_conv(input, future_context_size, param_attr=None, act=None):
    """Row Conv Operator. This layer will apply lookahead convolution to
    **input**. The input variable should be a 2D LoDTensor with shape [T, D].
    Parameters with shape [future_context_size + 1, D] will be created. The math
Y
yangyaming 已提交
3191
    equation of row convolution is as follows:
3192 3193 3194 3195 3196 3197 3198

    .. math::
        Out_{i} = \sum_{j = i} ^ {i + \\tau} X_{j} \odot W_{i - j}

    In the above equation:

    * :math:`Out_{i}`: The i-th row of output variable with shape [1, D].
Y
yangyaming 已提交
3199
    * :math:`\\tau`: Future context size.
3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
    * :math:`X_{j}`: The j-th row of input variable with shape [1, D].
    * :math:`W_{i-j}`: The (i-j)-th row of parameters with shape [1, D].

    More details about row_conv please refer to the paper \
    (http://www.cs.cmu.edu/~dyogatam/papers/wang+etal.iclrworkshop2016.pdf) and
    the design document \
    (https://github.com/PaddlePaddle/Paddle/issues/2228#issuecomment-303903645).

    Args:
        input (Variable): Input variable, a 2D LoDTensor with shape [T, D].
Y
yangyaming 已提交
3210 3211
        future_context_size (int): Future context size. Please note, the shape
            of convolution kernel is [future_context_size + 1, D].
3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236
        param_attr (ParamAttr): Attributes of parameters, including
            name, initializer etc.
        act (str): Non-linear activation to be applied to output variable.

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

    Examples:
        .. code-block:: python

            x = fluid.layers.data(name='x', shape=[16],
                            dtype='float32', lod_level=1)
            out = fluid.layers.row_conv(input=x, future_context_size=2)
    """
    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
3237
    return helper.append_activation(out)
3238 3239


3240 3241 3242 3243
def multiplex(inputs, index):
    """
    **Multiplex Layer**

Y
yangyaming 已提交
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258
    Referring to the given index variable, this layer selects rows from the
    input variables to construct a multiplex variable. Assuming that there are
    :math:`m` input variables and :math:`I_i` represents the i-th input
    variable and :math:`i` is in [0, :math:`m`). All input variables are
    tensors with same shape [:math:`d_0`, :math:`d_1`, ..., :math:`d_R`].
    Please note that rank of the input tensor should be at least 2. Each input
    variable will be treated as a 2-D matrix with shape [:math:`M`, :math:`N`]
    where :math:`M` for :math:`d_0` and :math:`N` for :math:`d_1` * :math:`d_2`
    * ... * :math:`d_R`. Let :math:`I_i[j]` be the j-th row of the i-th input
    variable. The given index variable should be a 2-D tensor with shape
    [:math:`M`, 1]. Let `ID[i]` be the i-th index value of the index variable.
    Then the output variable will be a tensor with shape [:math:`d_0`,
    :math:`d_1`, ..., :math:`d_R`]. If we treat the output tensor as a 2-D
    matrix with shape [:math:`M`, :math:`N`] and let :math:`O[i]` be the i-th
    row of the matrix, then `O[i]` is equal to :math:`I_{ID[i]}[i]`.
3259 3260

    Args:
Y
yangyaming 已提交
3261 3262
       inputs (list): A list of variables to gather from. All variables have the
                same shape and the rank is at least 2.
3263
       index (Variable): Tensor<int32>, index variable which is a 2-D tensor
Y
yangyaming 已提交
3264
                with shape [M, 1] where M is the batch size.
3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277

    Returns:
        Variable: Multiplex variable gathered from input variables.

    Examples:
        .. code-block:: python

            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)
    """
    helper = LayerHelper('multiplex', **locals())
Y
yangyaming 已提交
3278 3279 3280 3281 3282 3283

    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)
3284 3285 3286 3287 3288 3289
    helper.append_op(
        type='multiplex',
        inputs={'X': inputs,
                'Ids': index},
        outputs={'Out': [out]})
    return out
3290 3291 3292 3293 3294


def softmax_with_cross_entropy(logits, label, soft_label=False):
    """
    **Softmax With Cross Entropy Operator.**
3295

3296 3297 3298 3299
    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.
3300

3301 3302 3303
    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.
3304

3305 3306 3307
    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.
3308

3309
    The equation is as follows:
3310

3311
    1) Hard label (one-hot label, so every sample has exactly one class)
3312

3313 3314 3315 3316
    .. math::

        loss_j =  -\\text{logit}_{label_j} +
        \\log\\left(\\sum_{i=0}^{K}\\exp(\\text{logit}_i)\\right), j = 1,..., K
3317

3318 3319 3320
    2) Soft label (each sample can have a distribution over all classes)

    .. math::
3321

3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342
        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
3343 3344
            out = fluid.layers.softmax_with_cross_entropy(
                logits=fc, label=label)
3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362
    """
    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):
    """
    **Smooth L1 Loss Operator. **

Q
qingqing01 已提交
3363
    This operator computes the smooth L1 loss for X and Y.
3364
    The operator takes the first dimension of X and Y as batch size.
Q
qingqing01 已提交
3365
    For each instance, it computes the smooth L1 loss element by element first
3366
    and then sums all the losses. So the shape of Out is [batch_size, 1].
3367

3368 3369
    Args:
        x (Variable): A tensor with rank at least 2. The input value of smooth
Q
qingqing01 已提交
3370
            L1 loss op with shape [batch_size, dim1, ..., dimN].
3371
        y (Variable): A tensor with rank at least 2. The target value of smooth
Q
qingqing01 已提交
3372
            L1 loss op with same shape as x.
3373 3374 3375 3376 3377 3378
        inside_weight (Variable|None):  A tensor with rank at least 2. This
            input is optional and should have same shape with x. If provided,
            the result of (x - y) will be multiplied by this tensor element by
            element.
        outside_weight (Variable|None): A tensor with rank at least 2. This
            input is optional and should have same shape with x. If provided,
Q
qingqing01 已提交
3379
            the out smooth L1 loss will be multiplied by this tensor element
3380
            by element.
Q
qingqing01 已提交
3381
        sigma (float|None): Hyper parameter of smooth L1 loss op. A float scalar
3382 3383
            with default value 1.0.
    Returns:
Q
qingqing01 已提交
3384
        Variable: A tensor with rank be 2. The output smooth L1 loss with
3385 3386 3387 3388 3389 3390
            shape [batch_size, 1].

    Examples:
        .. code-block:: python

            data = fluid.layers.data(name='data', shape=[128], dtype='float32')
F
stash  
fengjiayi 已提交
3391 3392
            label = fluid.layers.data(
                name='label', shape=[100], dtype='float32')
3393
            fc = fluid.layers.fc(input=data, size=100)
F
fengjiayi 已提交
3394
            out = fluid.layers.smooth_l1(x=fc, y=label)
3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410
    """
    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
3411 3412 3413 3414 3415 3416 3417 3418 3419


def one_hot(input, depth):
    """
    One Hot Operator. This operator creates the one-hot representations for input
    index values. The following example will help to explain the function of this
    operator.

    Args:
F
fengjiayi 已提交
3420
        input(variable):  A Tensor/LodTensor of indices, last dimension must be 1.
3421 3422 3423 3424 3425 3426
        depth(scalar): an interger defining the depth of the one hot dimension.

    Returns:
         The one-hot tensor or LodTensor, same as input.

    Examples:
C
caoying03 已提交
3427 3428
        .. code-block:: python

3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449
        X is a LoDTensor:
          X.lod = [[0, 1, 4]]
          X.shape = [4, 1]
          X.data = [[1], [1], [3], [0]]
        set depth = 4
        Out is a LoDTensor:
          Out.lod = [[0, 1, 4]]
          Out.shape = [4, 4]
          Out.data = [[0., 1., 0., 0.],
                      [0., 1., 0., 0.],
                      [0., 0., 0., 1.],
                      [1., 0., 0., 0.]]
    """
    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
3450 3451


Y
Yu Yang 已提交
3452
def autoincreased_step_counter(counter_name=None, begin=1, step=1):
Y
Yu Yang 已提交
3453
    """
Y
Yu Yang 已提交
3454
    NOTE: The counter will be automatically increased by 1 every mini-batch
Y
Yu Yang 已提交
3455
    Return the run counter of the main program, which is started with 1.
Y
Yu Yang 已提交
3456 3457 3458 3459 3460 3461

    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.

Y
Yu Yang 已提交
3462 3463 3464
    Returns(Variable): The global run counter.
    """
    helper = LayerHelper('global_step_counter')
Y
Yu Yang 已提交
3465 3466
    if counter_name is None:
        counter_name = '@STEP_COUNTER@'
Y
Yu Yang 已提交
3467 3468 3469 3470 3471
    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
3472
                value=begin - 1, force_cpu=True))
Y
Yu Yang 已提交
3473 3474 3475
        helper.main_program.global_block().prepend_op(
            type='increment',
            inputs={'X': [counter]},
Y
Yu Yang 已提交
3476 3477
            outputs={'Out': [counter]},
            attrs={'step': float(step)})
Y
Yu Yang 已提交
3478 3479 3480
        counter.stop_gradient = True

    return counter
Y
yangyaming 已提交
3481 3482


3483
def reshape(x, shape, actual_shape=None, act=None, inplace=True, name=None):
C
caoying03 已提交
3484
    """
C
caoying03 已提交
3485 3486
    Gives a new shape to the input Tensor without changing its data.

3487 3488 3489 3490 3491
    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
3492

3493
    Some tricks exist when specifying the target shape.
C
caoying03 已提交
3494

3495 3496 3497 3498
    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.

3499
    2. 0 means the actual dimension value is going to be copied from the
3500 3501 3502 3503
    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
3504 3505

    1. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
W
wanghaoshuang 已提交
3506
    is [6, 8], the reshape operator will transform x into a 2-D tensor with
3507
    shape [6, 8] and leaving x's data unchanged.
C
caoying03 已提交
3508

3509
    2. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
3510 3511
    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
3512 3513
    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
3514
    dimensions.
C
caoying03 已提交
3515

3516
    3. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
3517 3518 3519 3520
    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
3521 3522 3523 3524 3525

    Args:
        input(variable): The input tensor.
        shape(list): The new shape. At most one dimension of the new shape can
                     be -1.
3526 3527 3528 3529 3530
        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
3531 3532 3533 3534 3535 3536 3537 3538 3539
        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.

    Returns(variable): The output tensor.

    Examples:
        .. code-block:: python
G
guosheng 已提交
3540

3541
            data = fluid.layers.data(
3542
                name='data', shape=[2, 4, 6], dtype='float32')
C
caoying03 已提交
3543
            reshaped = fluid.layers.reshape(
3544
                x=data, shape=[-1, 0, 3, 2], act='tanh', inplace=True)
C
caoying03 已提交
3545 3546 3547 3548 3549
    """

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

3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564
    # 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
3565 3566 3567 3568
    helper = LayerHelper("reshape", **locals())
    reshaped = helper.create_tmp_variable(dtype=x.dtype)
    helper.append_op(
        type="reshape",
3569 3570 3571
        inputs={"X": x,
                "Shape": actual_shape}
        if isinstance(actual_shape, Variable) else {"X": x},
C
caoying03 已提交
3572 3573 3574 3575 3576
        attrs={"shape": shape,
               "inplace": inplace},
        outputs={"Out": reshaped})

    return helper.append_activation(reshaped)
3577 3578


Y
yangyaming 已提交
3579
def lod_reset(x, y=None, target_lod=None):
Y
yangyaming 已提交
3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671
    """
    LoD Reset Operator. Set LoD of **x** to a new one specified by **y** or
    **target_lod**. When **y** provided, **y.lod** would be considered as target
    LoD first, otherwise **y.data** would be considered as target LoD. If **y**
    is not provided, target LoD should be specified by **target_lod**.
    If target LoD is specified by **Y.data** or **target_lod**, only one level
    LoD is supported.

    .. code-block:: text

        * Example 1:

            Given a 1-level LoDTensor x:
                x.lod =  [[ 0,     2,                   5      6 ]]
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

            target_lod: [0, 4, 6]

            then we get a 1-level LoDTensor:
                out.lod =  [[ 0,                   4,            6 ]]
                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:
                x.lod =  [[ 0,     2,                   5      6 ]]
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

            y is a Tensor:
                y.data = [[0, 2, 6]]
                y.dims = [1, 3]

            then we get a 1-level LoDTensor:
                out.lod =  [[ 0,     2,                          6 ]]
                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:
                x.lod =  [[ 0,      2,                   5     6 ]]
                x.data = [[1.0], [2.0], [3.0], [4.0], [5.0], [6.0]]
                x.dims = [6, 1]

            y is a 2-level LoDTensor:
                y.lod =  [[0, 2, 4], [0, 2, 5, 6]]
                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:
                out.lod =  [[0, 2, 4], [0, 2, 5, 6]]
                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 (Variable|None): If provided, output's LoD would be derived from y.
        target_lod (list|tuple|None): One level LoD which should be considered
                                      as target LoD when y not provided.

    Returns:
        Variable: Output variable with LoD specified by this operator.

    Raises:
        ValueError: If y and target_lod are both None.

    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
3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713


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::

        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}

    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
3714 3715
          data = fluid.layers.data(
              name="data", shape=[3, 112, 112], dtype="float32")
D
dragonwarrior 已提交
3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742
          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
3743 3744 3745 3746


def pad(x, paddings, pad_value=0., name=None):
    """
G
guosheng 已提交
3747
    Pads a tensor with a constant value given by :attr:`pad_value`, and the
W
wanghaoshuang 已提交
3748
    padded width is specified by :attr:`paddings`.
G
guosheng 已提交
3749

G
guosheng 已提交
3750 3751 3752 3753
    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
3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775

    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
3776
                         The length of :attr:paddings must be
G
guosheng 已提交
3777 3778 3779 3780 3781 3782 3783 3784 3785 3786
                         :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
3787

G
guosheng 已提交
3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801
            # 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
3802 3803 3804 3805 3806 3807 3808 3809 3810


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
3811 3812
    called label-smoothing regularization (LSR).

3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835
    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
3836
                              be :math:`(1, class\_num)`.
3837 3838
        epsilon(float): The weight used to mix up the original ground-truth
                        distribution and the fixed distribution.
3839
        dtype(np.dtype|core.VarDesc.VarType|str): The type of data : float32,
3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866
                                                  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
3867 3868 3869 3870


def roi_pool(input, rois, pooled_height=1, pooled_width=1, spatial_scale=1.0):
    """
3871
    Region of interest pooling (also known as RoI pooling) is to perform
3872 3873
        is to perform max pooling on inputs of nonuniform sizes to obtain
        fixed-size feature maps (e.g. 7*7).
3874 3875 3876 3877
    The operator has three steps:
        1. Dividing each region proposal into equal-sized sections with
           the pooled_width and pooled_height
        2. Finding the largest value in each section
3878 3879 3880 3881 3882 3883 3884
        3. Copying these max values to the output buffer

    Args:
        input (Variable): The input for ROI pooling.
        rois (Variable): ROIs (Regions of Interest) to pool over. It should
                         be a 2-D one level LoTensor of shape [num_rois, 4].
                         The layout is [x1, y1, x2, y2], where (x1, y1)
3885 3886
                         is the top left coordinates, and (x2, y2) is the
                         bottom right coordinates. The num_rois is the
3887 3888 3889 3890 3891 3892 3893 3894
                         total number of ROIs in this batch data.
        pooled_height (integer): The pooled output height. Default: 1
        pooled_width (integer): The pooled output width. Default: 1
        spatial_scale (float): Multiplicative spatial scale factor. To
                               translate ROI coords from their input scale
                               to the scale used when pooling. Default: 1.0

    Returns:
3895
        pool_out (Variable): The output is a 4-D tensor of the shape
3896 3897 3898
                             (num_rois, channels, pooled_h, pooled_w).

    Examples:
3899 3900
        .. code-block:: python

3901
            pool_out = fluid.layers.roi_pool(input=x, rois=rois, 7, 7, 1.0)
3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918
    """
    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
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946


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:
3947 3948
        .. code-block:: python

W
whs 已提交
3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
            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)
3960 3961


3962 3963 3964 3965 3966
def image_resize(input,
                 out_shape=None,
                 scale=None,
                 name=None,
                 resample='BILINEAR'):
3967
    """
3968
    Resize a batch of images.
F
stash  
fengjiayi 已提交
3969

3970 3971 3972 3973 3974
    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:
        'BILINEAR' : Bilinear interpolation
F
stash  
fengjiayi 已提交
3975

3976
    Args:
3977
        input (Variable): The input tensor of image resize layer,
3978 3979
                          This is a 4-D tensor of the shape
                          (num_batches, channels, in_h, in_w).
3980
        out_shape(list|tuple|Variable|None): Output shape of image resize
3981 3982
                                    layer, the shape is (out_h, out_w).
                                    Default: None
B
baiyf 已提交
3983
        scale(float|None): The multiplier for the input height or width.
3984 3985 3986
                         At least one of out_shape or scale must be set.
                         And out_shape has a higher priority than scale.
                         Default: None
3987 3988
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
3989 3990
        resample(str): The resample method. It can only be 'BILINEAR' currently.
                       Default: 'BILINEAR'
3991 3992 3993 3994

    Returns:
        out (Variable): The output is a 4-D tensor of the shape
                        (num_batches, channls, out_h, out_w).
F
stash  
fengjiayi 已提交
3995

3996 3997 3998
    Examples:
        .. code-block:: python

3999
            out = fluid.layers.image_resize(input, out_shape=[12, 12])
4000
    """
4001 4002 4003 4004
    resample_methods = {'BILINEAR': 'bilinear_interp'}
    if resample not in resample_methods:
        raise ValueError(
            "The 'resample' of image_resize can only be 'BILINEAR' currently.")
4005 4006
    if out_shape is None and scale is None:
        raise ValueError("One of out_shape and scale must not be None")
4007 4008
    helper = LayerHelper('bilinear_interp', **locals())
    dtype = helper.input_dtype()
4009 4010 4011 4012

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

4013 4014 4015
    out_h = 0
    out_w = 0
    inputs = {"X": input}
4016
    if out_shape is not None:
B
baiyf 已提交
4017 4018 4019
        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')
4020 4021 4022 4023 4024 4025
        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
4026 4027 4028 4029
    else:
        out_h = int(input.shape[2] * scale)
        out_w = int(input.shape[3] * scale)

4030 4031
    out = helper.create_tmp_variable(dtype)
    helper.append_op(
4032
        type=resample_methods[resample],
4033
        inputs=inputs,
4034 4035 4036 4037
        outputs={"Out": out},
        attrs={"out_h": out_h,
               "out_w": out_w})
    return out
F
stash  
fengjiayi 已提交
4038 4039


4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068
def resize_bilinear(input, out_shape=None, scale=None, name=None):
    """
    This is an alias of layer 'image_resize' with bilinear interpolation.

    The mathematical meaning of resize bilinear layer is
    Bilinear interpolation.
    Bilinear interpolation is an extension of linear interpolation for
    interpolating functions of two variables (e.g. H-direction and
    W-direction in this layer) on a rectilinear 2D grid.

    For details, please refer to Wikipedia:
    https://en.wikipedia.org/wiki/Bilinear_interpolation
    """

    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.
F
fengjiayi 已提交
4069

4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082
    Returns:
        out (Variable): The output is a 4-D tensor of the shape
                        (num_batches, channls, out_h, out_w).
    """
    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
4083 4084 4085
    out_shape[long_idx] = int(
        float(out_shape[long_idx]) * (float(out_short_len) / float(hw[
            short_idx])) + 0.5)
4086 4087 4088
    return image_resize(input=input, out_shape=out_shape, resample=resample)


W
whs 已提交
4089 4090 4091 4092 4093 4094 4095
def gather(input, index):
    """
    Output is obtained by gathering entries of the outer-most dimension 
    of X indexed by `index` and concatenate them together.

    .. math::

4096
        Out = X[Index]
W
whs 已提交
4097 4098 4099 4100 4101 4102 4103


    .. code-block:: text


                Given:

4104 4105
                X = [[1, 2],
                     [3, 4],
W
whs 已提交
4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137
                     [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:
        .. 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
4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156
@templatedoc()
def random_crop(x, shape, seed=None):
    """
    ${comment}

    Examples:
        >>> img = fluid.layers.data("img", [3, 256, 256])
        >>> cropped_img = fluid.layers.random_crop(img, shape=[3, 224, 224])

    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
stash  
fengjiayi 已提交
4157 4158 4159
    helper = LayerHelper("random_crop", **locals())
    dtype = helper.input_dtype()
    out = helper.create_tmp_variable(dtype)
Y
yuyang18 已提交
4160 4161 4162
    if seed is None:
        seed = random.randint(-65536, 65535)

F
stash  
fengjiayi 已提交
4163
    if isinstance(seed, int):
F
fengjiayi 已提交
4164
        seed_value = seed
F
fengjiayi 已提交
4165 4166 4167 4168 4169 4170 4171 4172
        seed = helper.create_tmp_variable(dtype="int64")
        helper.append_op(
            type="fill_constant",
            inputs={},
            outputs={"Out": seed},
            attrs={
                "dtype": seed.dtype,
                "shape": [1],
F
fengjiayi 已提交
4173 4174
                "value": float(seed_value),
                "force_cpu": True
F
fengjiayi 已提交
4175
            })
F
stash  
fengjiayi 已提交
4176 4177
    elif not isinstance(seed, Variable):
        raise ValueError("'seed' must be a Variable or an int.")
F
fengjiayi 已提交
4178
    seed_out = helper.create_tmp_variable(dtype="int64")
F
stash  
fengjiayi 已提交
4179 4180 4181 4182 4183 4184 4185 4186
    helper.append_op(
        type="random_crop",
        inputs={"X": input,
                "Seed": seed},
        outputs={"Out": out,
                 "SeedOut": seed_out},
        attrs={"shape": shape})
    return out