config_parser.py 130.8 KB
Newer Older
1
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
Z
zhangjinchao01 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101
#
# 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
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# 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.

from __future__ import print_function
'''
The following functions are available in the config file:

Bias: define bias. To be used as value of bias argument in Layer().

Data: define data provider.

Input: define input layer for a layer. To be used as element of inputs argument
       in Layer().

Conv: define a convolution operation for an input of a layer.

Norm: define a normalization operation for an input of a layer.

Pool: define a pooling operation for an input of a layer.

Layer: define a layer.

Parameter: define a parameter.

Import: import another config file. If the imported config file name is
        a relative path, then it will be searched under the directory of the
        current config file.

Inputs(layer_names...):
    Define the name of the input layers of the NeuralNetwork.
    The type of these layers must be "data".
    These layers will be provided with the DataBatch obtained
    from DataProvider. The data streams from DataProvider must
    have the same order.

Outputs(layer_names...):
    Define the name of the output layers of the NeuralNetwork.
    Usually the output is simply the cost layer.
    You can specify other layers as outputs and  calculate the
    cost (and its derivative) yourself.


default_initial_std(val)
default_initial_mean(val)
default_momentum(val):
default_decay_rate(val): Set the default value for these parameters


get_config_arg(name, type, default): Get the value for a config parameter.


*** customized extension to config_parser ***
The functionality of the config_parser can be extended.
If the config_arg_str for parse_config() contains
extension_module_name=[MODULE_NAME], then config_parser will call
MODULE_NAME.get_config_funcs(g_config)
MODULE_NAME.get_config_funcs() should return a dictionary of name to functions,
those functions will be available in the config file.
See trainer/tests/config_parser_test.py for example

To use this from paddle_trainer, paddle_trainer should be called with
--config_args=extension_module_name=[MODULE_NAME]

'''
import copy
import logging
import os
import sys
import traceback
import math
import shutil

try:
    from paddle.proto.DataConfig_pb2 import DataConfig
    from paddle.proto.ModelConfig_pb2 import ModelConfig
    from paddle.proto.ModelConfig_pb2 import LayerConfig
    from paddle.proto.ModelConfig_pb2 import LayerInputConfig
    from paddle.proto.ModelConfig_pb2 import ProjectionConfig
    from paddle.proto.ModelConfig_pb2 import OperatorConfig
    from paddle.proto.ModelConfig_pb2 import GeneratorConfig
    from paddle.proto.ModelConfig_pb2 import LinkConfig
    from paddle.proto.ParameterConfig_pb2 import ParameterConfig
    from paddle.proto.ParameterConfig_pb2 import ParameterUpdaterHookConfig
    from paddle.proto.TrainerConfig_pb2 import TrainerConfig

except Exception as e:
    traceback.print_exc()
    raise

logging.basicConfig(
Q
qijun 已提交
102
    format='[%(levelname)s %(asctime)s %(filename)s:%(lineno)s] %(message)s', )
Z
zhangjinchao01 已提交
103 104 105
logger = logging.getLogger('paddle')
logger.setLevel(logging.INFO)
__real_print__ = print
Q
qijun 已提交
106
print = logger.info
Z
zhangjinchao01 已提交
107 108 109 110

# from layer type name to layer class
g_layer_type_map = {}

Q
qijun 已提交
111

Z
zhangjinchao01 已提交
112 113 114
# Initialize global variables. We use this function so that we can
# call parse_config() multiple times
def init_config_environment(
Q
qijun 已提交
115 116 117 118 119 120 121 122 123 124 125 126 127 128
        g_default_momentum=None,
        g_default_decay_rate=None,
        g_default_initial_mean=0.,
        g_default_initial_std=0.01,
        g_default_num_batches_regularization=None,
        g_default_initial_strategy=0,
        g_default_initial_smart=False,
        g_default_gradient_clipping_threshold=None,
        g_default_device=None,
        g_default_update_hooks=None,
        g_default_compact_func=None,
        g_config=TrainerConfig(),
        g_layer_map={},
        g_parameter_map={},
X
xuwei06 已提交
129
        g_parameter_initializer_map={},
Q
qijun 已提交
130
        g_extended_config_funcs={},
Z
zhangjinchao01 已提交
131 132

        # store command args of paddle_trainer
Q
qijun 已提交
133
        g_command_config_args={},
Z
zhangjinchao01 已提交
134 135

        # Used for PyDataProvider to avoid duplicate module name
Q
qijun 已提交
136 137 138 139 140
        g_py_module_name_list=[],
        g_current_submodel=None,
        g_root_submodel=None,
        g_submodel_map={},
        g_submodel_stack=[],
141
        g_add_submodel_suffix=False, ):
Z
zhangjinchao01 已提交
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157

    for k, v in locals().iteritems():
        globals()[k] = copy.deepcopy(v)


# Because type is widely used as a variable name in this code.
# we need a different function name for the builtin type()
def type_of(x):
    return type(x)


# Check a condition derived config file
def config_assert(b, msg):
    if not b:
        logger.fatal(msg)

Q
qijun 已提交
158

Z
zhangjinchao01 已提交
159 160
g_config_funcs = {}

Q
qijun 已提交
161

Z
zhangjinchao01 已提交
162 163 164 165 166
# decorator for indicating a function which can be used in config file
def config_func(func):
    g_config_funcs[func.func_name] = func
    return func

Q
qijun 已提交
167

Z
zhangjinchao01 已提交
168 169 170 171 172
# decorator for indicating a class which can be used in config file
def config_class(cls):
    g_config_funcs[cls.__name__] = cls
    return cls

Q
qijun 已提交
173

Z
zhangjinchao01 已提交
174 175 176 177 178 179
# decorator for indicating a class for a layer type
def config_layer(layer_type):
    def wrap(cls):
        g_config_funcs[cls.__name__] = cls
        g_layer_type_map[layer_type] = cls
        return cls
Q
qijun 已提交
180

Z
zhangjinchao01 已提交
181 182
    return wrap

Q
qijun 已提交
183

Z
zhangjinchao01 已提交
184 185 186
def gen_parameter_name(layer_name, input_index):
    return '_%s.w%d' % (layer_name, input_index)

Q
qijun 已提交
187

Z
zhangjinchao01 已提交
188 189 190
def gen_bias_parameter_name(layer_name):
    return '_%s.wbias' % layer_name

Q
qijun 已提交
191

Z
zhangjinchao01 已提交
192 193 194
def default(x, default_value):
    return default_value if x is None else x

Q
qijun 已提交
195

Z
zhangjinchao01 已提交
196 197 198 199 200 201
class Cfg(object):
    def add_keys(self, locals):
        for k, v in locals.iteritems():
            if not k.startswith('_'):
                self.__setattr__(k, v)

Q
qijun 已提交
202

Z
zhangjinchao01 已提交
203 204
# functions available in config file

Q
qijun 已提交
205

Z
zhangjinchao01 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
# Define the name of the input layers of the NeuralNetwork.
# The type of these layers must be "data".
# These layers will be provided with the DataBatch obtained
# from DataProvider. The data streams from DataProvider must
# have the same order.
@config_func
def Inputs(*args):
    for name in args:
        name = MakeLayerNameInSubmodel(name)
        global g_current_submodel, g_root_submodel
        if g_current_submodel.is_recurrent_layer_group:
            config_assert(False, "Do not set Inputs in recurrent layer group")
        else:
            g_current_submodel.input_layer_names.append(name)

        if g_current_submodel is g_root_submodel:
            g_config.model_config.input_layer_names.append(name)

Q
qijun 已提交
224

225 226
@config_func
def HasInputsSet():
227
    return len(g_current_submodel.input_layer_names) != 0
228

Z
zhangjinchao01 已提交
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252

# Define the name of the output layers of the NeuralNetwork.
# Usually the output is simply the cost layer.
# You can specify other layers as outputs and calculate the
# cost (and its derivative) yourself.
@config_func
def Outputs(*args):
    for name in args:
        name = MakeLayerNameInSubmodel(name)
        global g_current_submodel, g_root_submodel
        if g_current_submodel.is_recurrent_layer_group:
            config_assert(False, "Do not set Outputs in recurrent layer group")
        else:
            g_current_submodel.output_layer_names.append(name)

        if g_current_submodel is g_root_submodel:
            g_config.model_config.output_layer_names.append(name)


@config_func
def SubModelBegin(name):
    global g_current_submodel, g_root_submodel, g_submodel_stack
    g_submodel_stack.append(g_current_submodel)

Q
qijun 已提交
253
    name = MakeLayerNameInParentSubmodel(name)  #rename in nested submodel
Z
zhangjinchao01 已提交
254 255 256 257 258 259 260 261 262

    config_assert(name not in g_submodel_map,
                  'Duplicated submodel name: %s' % name)

    sub_model = g_config.model_config.sub_models.add()
    sub_model.name = name
    g_submodel_map[name] = sub_model
    g_current_submodel = sub_model

Q
qijun 已提交
263

Z
zhangjinchao01 已提交
264
@config_func
Q
qijun 已提交
265
def SubModelEnd(name=None):
Z
zhangjinchao01 已提交
266
    global g_current_submodel, g_root_submodel, g_submodel_stack
Q
qijun 已提交
267 268
    config_assert(g_current_submodel is not g_root_submodel,
                  "submodel not begin")
Z
zhangjinchao01 已提交
269
    if name is not None:
Q
qijun 已提交
270 271 272
        config_assert(
            g_current_submodel.name == MakeLayerNameInParentSubmodel(name),
            "submodel name error")
Z
zhangjinchao01 已提交
273 274 275

    g_current_submodel = g_submodel_stack.pop()

Q
qijun 已提交
276

Z
zhangjinchao01 已提交
277 278
def MakeLayerNameInParentSubmodel(name):
    suffix = ""
279 280
    if len(g_submodel_stack) > 1:
        suffix = "@" + g_submodel_stack[-1].name
Z
zhangjinchao01 已提交
281 282
    return name + suffix

Q
qijun 已提交
283

Z
zhangjinchao01 已提交
284 285 286
def GetLayerBaseName(name):
    return name.split('@')[0]

Q
qijun 已提交
287 288

def MakeLayerNameInSubmodel(name, submodel_name=None):
Z
zhangjinchao01 已提交
289 290
    global g_current_submodel
    global g_add_submodel_suffix
Q
qijun 已提交
291 292
    if (submodel_name is None and not g_add_submodel_suffix and
            not g_current_submodel.is_recurrent_layer_group):
Z
zhangjinchao01 已提交
293 294 295 296 297
        return name
    if submodel_name is None:
        submodel_name = g_current_submodel.name
    return name + "@" + submodel_name

Q
qijun 已提交
298

Z
zhangjinchao01 已提交
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
# Define a recurrent layer group begin with RecurrentLayerGroupBegin
# and end with RecurrentLayerGroupEnd.
# A recurrent layer group forward/backward one frame after previous frame
# forward/backward through all layers in layer group.
# in_links are names of layer used as input layer in the layer group.
# out_links are names of layer in layer group used as outside layer's input.
#
# If generator is set, the layer group need one or more than one outlinks.
# The first outlink should always be the generated token ids.
# If generator.num_results_per_sample is not set, the output for one sample is
# a ids sequence. Else if num_results_per_sample is more than one,
# the output for one sample is up to #num_results_per_sample generated
# sequences, which are packed in one sequence in output ids vector. Each
# generated sequence has a generation probability. The probabilities for one
# sample are stored in one row of output value matrix.
# Packed generated sequences format, for each i:
#   seq_i_length: one interger, seq_i content length,
#   [seq_i content], length = seq_i_length
#   seq_i_end_mark: one interger, for format check, always -1
# You can use "seq_text_printer" to print the output of the generator.
@config_func
def RecurrentLayerGroupWithoutOutLinksBegin(name,
                                            in_links,
322 323
                                            seq_reversed=False,
                                            target_inlinkname=""):
Z
zhangjinchao01 已提交
324 325 326 327 328 329 330 331
    global g_current_submodel
    config_assert(g_config.model_config.type == "recurrent_nn",
                  "RecurrentLayerGroup should be used only in recurrent_nn")
    RecurrentLayerGroup(name=name)  # add to father model
    SubModelBegin(name)
    g_current_submodel.is_recurrent_layer_group = True
    g_current_submodel.reversed = seq_reversed
    in_links_count = 0
332
    for linkid, link in enumerate(in_links):
Z
zhangjinchao01 已提交
333 334 335 336
        if isinstance(link, basestring):
            name = link
        else:
            name = link.link_name
337

Z
zhangjinchao01 已提交
338 339 340
        in_links_count += 1
        layer_name = MakeLayerNameInParentSubmodel(name)
        layer = g_layer_map[layer_name]
341
        ScatterAgentLayer(name=name, size=layer.size)
342

Z
zhangjinchao01 已提交
343 344 345 346
        pair = g_current_submodel.in_links.add()
        pair.layer_name = layer_name
        pair.link_name = MakeLayerNameInSubmodel(name)

Q
qijun 已提交
347

Z
zhangjinchao01 已提交
348 349 350 351 352 353 354 355 356 357 358 359 360
@config_func
def RecurrentLayerGroupSetOutLink(link):
    if isinstance(link, basestring):
        name = link
    else:
        name = link.link_name
    layer_name = MakeLayerNameInParentSubmodel(name)
    pair = g_current_submodel.out_links.add()
    pair.layer_name = MakeLayerNameInSubmodel(name)
    pair.link_name = layer_name


def RecurrentLayerGroupSetGenerator(generator=None):
Q
qijun 已提交
361
    generator.eos_layer_name = MakeLayerNameInSubmodel(generator.eos_layer_name)
Z
zhangjinchao01 已提交
362 363 364 365 366 367 368 369
    g_current_submodel.generator.CopyFrom(generator)


@config_func
def RecurrentLayerGroupBegin(name,
                             in_links,
                             out_links,
                             generator=None,
370
                             target_inlinkname="",
Z
zhangjinchao01 已提交
371
                             seq_reversed=False):
372
    RecurrentLayerGroupWithoutOutLinksBegin(name, in_links, seq_reversed)
Z
zhangjinchao01 已提交
373 374 375 376 377
    for link in out_links:
        RecurrentLayerGroupSetOutLink(link)

    if generator is not None:
        RecurrentLayerGroupSetGenerator(generator)
Q
qijun 已提交
378 379 380 381 382
        config_assert(
            len(in_links) == 0, "no in_links should be passed to generator")
        config_assert(
            len(out_links) >= 1,
            "one or more than one out_links should be passed to generator")
Z
zhangjinchao01 已提交
383 384 385 386 387 388 389


@config_func
def RecurrentLayerGroupEnd(name):
    global g_current_submodel
    config_assert(g_current_submodel.is_recurrent_layer_group,
                  "RecurrentLayerGroup not begin")
Q
qijun 已提交
390
    for pair in g_current_submodel.memories:  #check exist
Z
zhangjinchao01 已提交
391
        layer = g_layer_map[pair.layer_name]
Y
Yu Yang 已提交
392 393
        config_assert(layer is not None,
                      "memory declare wrong name:%s" % pair.layer_name)
Z
zhangjinchao01 已提交
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
        memory_link = g_layer_map[pair.link_name]
        config_assert(layer.size == memory_link.size,
                      "memory declare wrong size:%d" % memory_link.size)

    prev_submodel = g_current_submodel
    SubModelEnd(name)

    for pair in prev_submodel.out_links:
        layer = g_layer_map[pair.layer_name]
        # add out agent to father model
        agent_name = GetLayerBaseName(pair.link_name)
        if prev_submodel.HasField("generator"):
            DataLayer(name=agent_name, size=layer.size)
        else:
            GatherAgentLayer(name=agent_name, size=layer.size)

Q
qijun 已提交
410

Z
zhangjinchao01 已提交
411 412 413 414 415 416
# Define the model type
# currently, the paddle supports "nn", "recurrent_nn", "recursive_nn" and "multi_nn"
@config_func
def model_type(name):
    g_config.model_config.type = name

Q
qijun 已提交
417

Z
zhangjinchao01 已提交
418 419
@config_class
class Bias(Cfg):
X
xuwei06 已提交
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435
    def __init__(self,
                 parameter_name=None,
                 learning_rate=None,
                 momentum=None,
                 decay_rate=None,
                 decay_rate_l1=None,
                 initial_mean=None,
                 initial_std=None,
                 initial_strategy=None,
                 initial_smart=None,
                 num_batches_regularization=None,
                 sparse_remote_update=None,
                 gradient_clipping_threshold=None,
                 is_static=None,
                 is_shared=None,
                 initializer=None):
Z
zhangjinchao01 已提交
436 437
        self.add_keys(locals())

Q
qijun 已提交
438

Z
zhangjinchao01 已提交
439 440 441 442 443 444 445
# Define one input for a layer
@config_class
class Input(Cfg):
    def __init__(
            self,
            input_layer_name,
            parameter_name=None,
X
xuwei06 已提交
446
            initializer=None,
Z
zhangjinchao01 已提交
447 448 449 450 451 452 453 454 455 456 457 458 459
            learning_rate=None,
            momentum=None,
            decay_rate=None,
            decay_rate_l1=None,
            initial_mean=None,
            initial_std=None,
            initial_strategy=None,
            initial_smart=None,
            num_batches_regularization=None,
            sparse_remote_update=None,
            sparse_update=None,
            gradient_clipping_threshold=None,
            conv=None,
L
liaogang 已提交
460
            bilinear_interp=None,
Z
zhangjinchao01 已提交
461 462 463 464
            norm=None,
            pool=None,
            image=None,
            block_expand=None,
465
            maxout=None,
Q
qijun 已提交
466
            spp=None,
D
dangqingqing 已提交
467
            pad=None,
Z
zhangjinchao01 已提交
468 469 470 471 472
            format=None,
            nnz=None,
            is_static=None,
            is_shared=None,
            update_hooks=None,
473
            input_layer_argument=None,
D
dangqingqing 已提交
474 475 476 477 478
            make_layer_name_in_submodel=True, ):
        """
        @param make_layer_name_in_submodel True by defalut, you might need to
        set it carefully when adding Input in config_parser.py.
        """
Z
zhangjinchao01 已提交
479
        self.add_keys(locals())
D
dangqingqing 已提交
480 481 482
        self.input_layer_name = MakeLayerNameInSubmodel(
            input_layer_name
        ) if make_layer_name_in_submodel else input_layer_name
Z
zhangjinchao01 已提交
483

Q
qijun 已提交
484

Z
zhangjinchao01 已提交
485 486 487
# Define a projection for iexed layer
@config_class
class Projection(Input):
Q
qijun 已提交
488 489
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
490 491 492
    def __init__(
            self,
            input_layer_name,
Q
qijun 已提交
493
            size=0,  # projection output size
Z
zhangjinchao01 已提交
494 495 496 497 498 499 500 501 502
            parameter_name=None,
            learning_rate=None,
            momentum=None,
            decay_rate=None,
            decay_rate_l1=None,
            initial_mean=None,
            initial_std=None,
            initial_strategy=None,
            initial_smart=None,
X
xuwei06 已提交
503
            initializer=None,
Z
zhangjinchao01 已提交
504 505 506 507 508 509 510 511 512 513
            num_batches_regularization=None,
            sparse_remote_update=None,
            sparse_update=None,
            gradient_clipping_threshold=None,
            ptype=None,
            format=None,
            nnz=None,
            is_static=None,
            is_shared=None,
            update_hooks=None,
Q
qijun 已提交
514
            input_layer_argument=None, ):
Z
zhangjinchao01 已提交
515 516 517 518 519 520 521 522 523 524 525 526 527
        self.add_keys(locals())
        self.input_layer_name = MakeLayerNameInSubmodel(input_layer_name)

        self.proj_conf = ProjectionConfig()
        if ptype is not None:
            self.proj_conf.type = ptype
        else:
            self.proj_conf.type = self.type

    # calculate the output_size given input_size. return 0
    # to indicate using the size from Layer config
    def calc_output_size(self, input_layer_config):
        return self.size
Q
qijun 已提交
528

Z
zhangjinchao01 已提交
529 530
    def calc_parameter_size(self, input_size, output_size):
        raise NotimplementedError
Q
qijun 已提交
531

Z
zhangjinchao01 已提交
532 533 534 535 536 537 538 539 540 541
    def calc_parameter_dims(self, input_size, output_size):
        raise NotimplementedError


@config_class
class IdentityProjection(Projection):
    type = 'identity'

    def calc_output_size(self, input_layer_config):
        return input_layer_config.size
Q
qijun 已提交
542

Z
zhangjinchao01 已提交
543 544
    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
545

Z
zhangjinchao01 已提交
546 547 548
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
549

Z
zhangjinchao01 已提交
550 551 552 553 554 555
# Like IdentityProjection, but layer size may smaller than input size,
# the projection select dimesions [offset, offset+layer_size) from input
@config_class
class IdentityOffsetProjection(Projection):
    type = 'identity_offset'

Q
qijun 已提交
556 557 558
    def __init__(self, input_layer_name, offset, **xargs):
        super(IdentityOffsetProjection, self).__init__(input_layer_name,
                                                       **xargs)
Z
zhangjinchao01 已提交
559 560 561 562
        self.proj_conf.offset = offset

    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
563

Z
zhangjinchao01 已提交
564 565 566
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
567

Z
zhangjinchao01 已提交
568 569 570 571 572 573 574
# DotMulProjection performs element-wise multiplication with weight
@config_class
class DotMulProjection(Projection):
    type = 'dot_mul'

    def calc_output_size(self, input_layer_config):
        return input_layer_config.size
Q
qijun 已提交
575

Z
zhangjinchao01 已提交
576 577
    def calc_parameter_size(self, input_size, output_size):
        return output_size
Q
qijun 已提交
578

Z
zhangjinchao01 已提交
579 580 581
    def calc_parameter_dims(self, input_size, output_size):
        return [1, output_size]

L
Luo Tao 已提交
582

X
xuwei06 已提交
583 584 585 586 587 588 589 590 591 592 593 594 595 596
# ScalingProjection
@config_class
class ScalingProjection(Projection):
    type = 'scaling'

    def calc_output_size(self, input_layer_config):
        return input_layer_config.size

    def calc_parameter_size(self, input_size, output_size):
        return 1

    def calc_parameter_dims(self, input_size, output_size):
        return [1, 1]

Q
qijun 已提交
597

Z
zhangjinchao01 已提交
598 599 600 601 602 603
@config_class
class TableProjection(Projection):
    type = 'table'

    def calc_parameter_size(self, input_size, output_size):
        return input_size * output_size
Q
qijun 已提交
604

Z
zhangjinchao01 已提交
605 606 607
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
608

Z
zhangjinchao01 已提交
609 610 611 612 613 614
@config_class
class FullMatrixProjection(Projection):
    type = 'fc'

    def calc_parameter_size(self, input_size, output_size):
        return input_size * output_size
Q
qijun 已提交
615

Z
zhangjinchao01 已提交
616 617 618
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
619

Z
zhangjinchao01 已提交
620 621 622 623 624 625
@config_class
class TransposedFullMatrixProjection(Projection):
    type = 'trans_fc'

    def calc_parameter_size(self, input_size, output_size):
        return input_size * output_size
Q
qijun 已提交
626

Z
zhangjinchao01 已提交
627 628 629
    def calc_parameter_dims(self, input_size, output_size):
        return [output_size, input_size]

Q
qijun 已提交
630

Z
zhangjinchao01 已提交
631 632 633 634
@config_class
class ContextProjection(Projection):
    type = 'context'

Q
qijun 已提交
635 636
    def __init__(self, input_layer_name, context_start, context_length,
                 trainable_padding, **xargs):
Z
zhangjinchao01 已提交
637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659
        super(ContextProjection, self).__init__(input_layer_name, **xargs)
        self.proj_conf.context_start = context_start
        self.proj_conf.context_length = context_length
        self.proj_conf.trainable_padding = trainable_padding
        self._total_pad = max(0, -self.proj_conf.context_start) \
                          + max(0, self.proj_conf.context_start \
                                + self.proj_conf.context_length - 1)

    def calc_output_size(self, input_layer_config):
        return input_layer_config.size * self.proj_conf.context_length

    def calc_parameter_size(self, input_size, output_size):
        if self.proj_conf.trainable_padding == False:
            return 0
        else:
            return input_size * self._total_pad

    def calc_parameter_dims(self, input_size, output_size):
        return [self._total_pad, input_size]

    _total_pad = 0


660
@config_class
661
class ConvBaseProjection(Projection):
Q
qijun 已提交
662 663 664 665 666
    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
667
        super(ConvBaseProjection, self).__init__(input_layer_name, **xargs)
668 669 670 671 672 673 674 675 676 677 678 679

        if num_filters is not None:
            self.proj_conf.num_filters = num_filters

    def calc_output_size(self, input_layer_config):
        return self.proj_conf.output_size

    def calc_parameter_size(self, input_size, output_size):
        co = self.proj_conf.num_filters
        ci = self.proj_conf.conv_conf.channels
        fh = self.proj_conf.conv_conf.filter_size
        fw = self.proj_conf.conv_conf.filter_size_y
680 681
        gr = self.proj_conf.conv_conf.groups
        return co * ci * fh * fw / gr
682 683 684 685 686 687 688

    def calc_bias_size(self):
        return self.proj_conf.num_filters

    def calc_parameter_dims(self, input_size, output_size):
        return None

Q
qijun 已提交
689

690 691 692 693 694 695 696 697 698
@config_class
class ConvProjection(ConvBaseProjection):
    type = 'conv'

    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
699 700
        super(ConvProjection, self).__init__(input_layer_name, num_filters,
                                             conv_conf, **xargs)
701

702
        parse_conv(conv_conf, self.input_layer_name, self.proj_conf.conv_conf,
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
                   num_filters)
        self.proj_conf.output_size = self.proj_conf.conv_conf.output_x * \
                                     self.proj_conf.conv_conf.output_y * \
                                     num_filters


@config_class
class ConvTransProjection(ConvBaseProjection):
    type = 'convt'

    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
718 719
        super(ConvTransProjection, self).__init__(input_layer_name, num_filters,
                                                  conv_conf, **xargs)
720 721 722

        parse_conv(
            conv_conf,
723
            self.input_layer_name,
724 725 726 727 728 729 730 731
            self.proj_conf.conv_conf,
            num_filters,
            trans=True)
        self.proj_conf.output_size = self.proj_conf.conv_conf.img_size_y * \
                                     self.proj_conf.conv_conf.img_size * \
                                     num_filters


Z
zhangjinchao01 已提交
732 733 734
# Define a operator for mixed layer
@config_class
class Operator(Cfg):
Q
qijun 已提交
735 736
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
737 738
    def __init__(
            self,
Q
qijun 已提交
739
            input_layer_names, ):
Z
zhangjinchao01 已提交
740 741 742 743 744 745 746 747 748 749
        self.add_keys(locals())
        self.operator_conf = OperatorConfig()
        self.operator_conf.type = self.type

    def check_dims(self):
        pass

    def calc_output_size(self, input_sizes):
        return 0

Q
qijun 已提交
750

Z
zhangjinchao01 已提交
751 752 753
@config_class
class DotMulOperator(Operator):
    type = 'dot_mul'
Q
qijun 已提交
754 755 756

    def __init__(self, input_layer_names, scale=None, **xargs):
        super(DotMulOperator, self).__init__(input_layer_names, **xargs)
Z
zhangjinchao01 已提交
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
        if scale is not None:
            self.operator_conf.dotmul_scale = scale

        config_assert(len(input_layer_names) == 2, "DotMul is binary operator")

    def check_dims(self):
        for i in range(2):
            config_assert(self.operator_conf.input_sizes[i] ==
                          self.operator_conf.output_size,
                          "DotMul input_size != output_size")

    def calc_output_size(self, input_sizes):
        return input_sizes[0]


@config_class
class ConvOperator(Operator):
    type = 'conv'
Q
qijun 已提交
775 776 777 778 779 780 781

    def __init__(self,
                 input_layer_names,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
        super(ConvOperator, self).__init__(input_layer_names, **xargs)
Z
zhangjinchao01 已提交
782 783 784
        if num_filters is not None:
            self.operator_conf.num_filters = num_filters

785 786
        parse_conv(conv_conf,
                   MakeLayerNameInSubmodel(input_layer_names[0]),
Q
qijun 已提交
787
                   self.operator_conf.conv_conf, num_filters)
L
Luo Tao 已提交
788 789 790
        self.operator_conf.output_size = self.operator_conf.conv_conf.output_x * \
                                         self.operator_conf.conv_conf.output_y * \
                                         num_filters
Z
zhangjinchao01 已提交
791 792 793

        config_assert(len(input_layer_names) == 2, "Conv is binary operator")

794 795
    def calc_output_size(self, input_sizes):
        return self.operator_conf.output_size
Z
zhangjinchao01 已提交
796 797


798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
@config_class
class ConvTransOperator(Operator):
    type = 'convt'

    def __init__(self,
                 input_layer_names,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
        super(ConvTransOperator, self).__init__(input_layer_names, **xargs)
        if num_filters is not None:
            self.operator_conf.num_filters = num_filters

        parse_conv(
            conv_conf,
            MakeLayerNameInSubmodel(input_layer_names[0]),
            self.operator_conf.conv_conf,
            num_filters,
            trans=True)
        self.operator_conf.output_size = \
            self.operator_conf.conv_conf.img_size * \
            self.operator_conf.conv_conf.img_size_y * \
            num_filters

        config_assert(len(input_layer_names) == 2, "Conv is binary operator")

    def calc_output_size(self, input_sizes):
        return self.operator_conf.output_size


Z
zhangjinchao01 已提交
828 829 830
# please refer to the comments in proto/ModelConfig.proto
@config_class
class Conv(Cfg):
Q
qijun 已提交
831 832 833 834 835 836 837 838 839 840 841 842 843
    def __init__(self,
                 filter_size,
                 channels,
                 padding=None,
                 stride=None,
                 groups=None,
                 filter_channels=None,
                 output_x=None,
                 img_size=None,
                 caffe_mode=True,
                 filter_size_y=None,
                 padding_y=None,
                 stride_y=None):
Z
zhangjinchao01 已提交
844 845
        self.add_keys(locals())
        if filter_size_y is None:
Q
qijun 已提交
846
            self.filter_size_y = filter_size
Z
zhangjinchao01 已提交
847
        if padding_y is None:
Q
qijun 已提交
848
            self.padding_y = padding
Z
zhangjinchao01 已提交
849
        if stride_y is None:
Q
qijun 已提交
850
            self.stride_y = stride
Z
zhangjinchao01 已提交
851
        if output_x is not None:
Q
qijun 已提交
852 853
            config_assert(output_x <= 0)

Z
zhangjinchao01 已提交
854

L
liaogang 已提交
855 856
@config_class
class BilinearInterp(Cfg):
L
Luo Tao 已提交
857
    def __init__(self, out_size_x=None, out_size_y=None, channels=None):
L
liaogang 已提交
858 859
        self.add_keys(locals())

Q
qijun 已提交
860

Z
zhangjinchao01 已提交
861 862
@config_class
class Pool(Cfg):
D
dangqingqing 已提交
863 864 865 866 867 868 869 870 871 872 873
    def __init__(
            self,
            pool_type,
            channels,
            size_x,
            size_y=None,
            start=None,
            stride=None,  # 1 by defalut in protobuf
            stride_y=None,
            padding=None,  # 0 by defalut in protobuf
            padding_y=None):
Z
zhangjinchao01 已提交
874
        self.add_keys(locals())
Q
qijun 已提交
875 876


Q
qijun 已提交
877
@config_class
Q
qijun 已提交
878
class SpatialPyramidPool(Cfg):
L
Luo Tao 已提交
879
    def __init__(self, pool_type, pyramid_height, channels):
Q
qijun 已提交
880
        self.add_keys(locals())
Z
zhangjinchao01 已提交
881

Q
qijun 已提交
882

D
dangqingqing 已提交
883 884 885 886 887 888
@config_class
class Pad(Cfg):
    def __init__(self, channels, pad_c, pad_h, pad_w):
        self.add_keys(locals())


Z
zhangjinchao01 已提交
889 890
@config_class
class Norm(Cfg):
Q
qijun 已提交
891 892 893 894 895 896 897 898 899
    def __init__(self,
                 norm_type,
                 channels,
                 size,
                 scale,
                 pow,
                 output_x=None,
                 img_size=None,
                 blocked=None):
Z
zhangjinchao01 已提交
900 901
        self.add_keys(locals())

Q
qijun 已提交
902

Z
zhangjinchao01 已提交
903 904
@config_class
class Image(Cfg):
Q
qijun 已提交
905
    def __init__(self, channels, img_size=None):
Z
zhangjinchao01 已提交
906 907
        self.add_keys(locals())

Q
qijun 已提交
908

Z
zhangjinchao01 已提交
909 910
@config_class
class BlockExpand(Cfg):
Q
qijun 已提交
911 912 913 914 915 916 917 918 919 920 921 922
    def __init__(self,
                 channels,
                 padding_x=0,
                 padding_y=0,
                 stride_x=0,
                 stride_y=0,
                 block_x=0,
                 block_y=0,
                 img_size_x=0,
                 img_size_y=0,
                 output_x=0,
                 output_y=0):
Z
zhangjinchao01 已提交
923 924
        self.add_keys(locals())

Q
qijun 已提交
925

926 927
@config_class
class MaxOut(Cfg):
Q
qijun 已提交
928
    def __init__(self, channels, groups, img_size_x=0, img_size_y=0):
929 930
        self.add_keys(locals())

Q
qijun 已提交
931

932
def create_data_config_proto(async_load_data=False,
933
                             constant_slots=None,
王益 已提交
934 935 936
                             data_ratio=1,
                             is_main_data=True,
                             usage_ratio=None):
Z
zhangjinchao01 已提交
937 938 939 940 941 942 943 944
    # default: all sub dataproviders are treat as "main data".
    # see proto/DataConfig.proto for is_main_data
    data_config = DataConfig()

    data_config.async_load_data = async_load_data

    if constant_slots:
        data_config.constant_slots.extend(constant_slots)
Q
qijun 已提交
945 946
    data_config.data_ratio = data_ratio
    data_config.is_main_data = is_main_data
Z
zhangjinchao01 已提交
947

Q
qijun 已提交
948
    usage_ratio = default(usage_ratio, settings_deprecated["usage_ratio"])
Z
zhangjinchao01 已提交
949 950 951 952 953 954
    config_assert(usage_ratio >= 0 and usage_ratio <= 1,
                  "The range of usage_ratio is [0, 1]")
    data_config.usage_ratio = usage_ratio

    return data_config

Q
qijun 已提交
955

Z
zhangjinchao01 已提交
956
@config_func
Q
qijun 已提交
957 958 959 960 961
def SimpleData(files=None,
               feat_dim=None,
               context_len=None,
               buffer_capacity=None,
               **xargs):
962
    data_config = create_data_config_proto(**xargs)
Z
zhangjinchao01 已提交
963 964 965 966 967 968 969 970 971
    data_config.type = 'simple'
    data_config.files = files
    data_config.feat_dim = feat_dim
    if context_len is not None:
        data_config.context_len = context_len
    if buffer_capacity:
        data_config.buffer_capacity = buffer_capacity
    return data_config

Q
qijun 已提交
972

Z
zhangjinchao01 已提交
973
@config_func
Q
qijun 已提交
974 975 976 977 978 979 980 981 982 983
def PyData(files=None,
           type=None,
           file_group_queue_capacity=None,
           load_data_module=None,
           load_data_object=None,
           load_data_args="",
           load_file_count=None,
           constant_slots=None,
           load_thread_num=None,
           **xargs):
984
    data_config = create_data_config_proto(**xargs)
Z
zhangjinchao01 已提交
985 986
    data_config.type = 'py'
    if load_data_module in g_py_module_name_list:
Q
qijun 已提交
987

Z
zhangjinchao01 已提交
988 989 990
        def get_path(module):
            m = __import__(load_data_module)
            return os.path.split(os.path.realpath(m.__file__))[0]
Q
qijun 已提交
991

Z
zhangjinchao01 已提交
992 993 994
        # python C-api is not thread safe, one module can only be import once,
        # so here we nedd to copy the module with different names if it has to be
        # imported several times.
Q
qijun 已提交
995 996
        module_new_name = "%s_copy_%d" % (load_data_module,
                                          len(g_py_module_name_list))
Z
zhangjinchao01 已提交
997
        g_py_module_name_list.append(module_new_name)
Q
qijun 已提交
998 999 1000 1001
        module_path = "%s/%s.py" % (get_path(load_data_module),
                                    load_data_module)
        new_module_path = "%s/%s.py" % (get_path(load_data_module),
                                        module_new_name)
Z
zhangjinchao01 已提交
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
        if os.path.isfile(module_path) == False:
            raise Exception("File %s is not exist." % module_path)
        shutil.copy2(module_path, new_module_path)
        load_data_module = module_new_name
    else:
        g_py_module_name_list.append(load_data_module)
    if load_data_module is not None and load_data_object is not None:
        data_config.load_data_module = load_data_module
        data_config.load_data_object = load_data_object
    else:
        raise ValueError('load_data_module, load_data_object is not defined.')
    data_config.load_data_args = load_data_args

    data_config.files = files or ''
    if file_group_queue_capacity is not None:
        data_config.file_group_conf.queue_capacity = file_group_queue_capacity
    if load_file_count is not None:
        data_config.file_group_conf.load_file_count = load_file_count
    if load_thread_num is not None:
        data_config.file_group_conf.load_thread_num = load_thread_num
    if constant_slots:
        data_config.constant_slots.extend(constant_slots)
    return data_config

Q
qijun 已提交
1026

Z
zhangjinchao01 已提交
1027
@config_func
Q
qijun 已提交
1028 1029 1030 1031 1032 1033 1034
def ProtoData(files=None,
              type=None,
              file_group_queue_capacity=None,
              load_file_count=None,
              constant_slots=None,
              load_thread_num=None,
              **xargs):
1035
    data_config = create_data_config_proto(**xargs)
Z
zhangjinchao01 已提交
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
    if type is None:
        data_config.type = 'proto'
    else:
        data_config.type = type
    data_config.files = files

    # When type="proto_group", one data provider contains at most
    # load_file_count files, and there are at most
    # (queue_capacity + load_thread_num + 1) data providers in memory
    if file_group_queue_capacity is not None:
        data_config.file_group_conf.queue_capacity = file_group_queue_capacity
    if load_file_count is not None:
        data_config.file_group_conf.load_file_count = load_file_count
    if load_thread_num is not None:
        data_config.file_group_conf.load_thread_num = load_thread_num
    if constant_slots:
        data_config.constant_slots.extend(constant_slots)
    return data_config

Q
qijun 已提交
1055

Z
zhangjinchao01 已提交
1056 1057
#real data for training is actually provided by "sub_data" data providers.
@config_func
Q
qijun 已提交
1058
def MultiData(sub_data=[]):
Z
zhangjinchao01 已提交
1059 1060 1061 1062 1063
    data_config = DataConfig()
    data_config.type = 'multi'
    data_config.sub_data_configs.extend(sub_data)
    return data_config

Q
qijun 已提交
1064

Z
zhangjinchao01 已提交
1065
@config_func
Q
qijun 已提交
1066 1067 1068 1069 1070 1071 1072
def Data(type,
         files=None,
         feat_dim=None,
         slot_dims=None,
         context_len=None,
         buffer_capacity=None,
         **xargs):
Z
zhangjinchao01 已提交
1073

1074
    data_config = create_data_config_proto(**xargs)
Z
zhangjinchao01 已提交
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
    data_config.type = type
    data_config.files = files
    data_config.feat_dim = feat_dim
    data_config.slot_dims.extend(slot_dims)
    if context_len is not None:
        data_config.context_len = context_len
    data_config.buffer_capacity = buffer_capacity
    return data_config


@config_func
def TrainData(data_config, async_load_data=None):
    config_assert(not g_config.HasField('data_config'),
                  'Only one TrainData definition is allowed')
    g_config.data_config.CopyFrom(data_config)
    g_config.data_config.for_test = False
    if async_load_data is not None:
        logger.warning("Deprecated: async_load_data should be used inside"
                       " Data definition")
        g_config.data_config.async_load_data = async_load_data


@config_func
def TestData(data_config, async_load_data=None):
    config_assert(not g_config.HasField('test_data_config'),
                  'Only one TestData definition is allowed')
    g_config.test_data_config.CopyFrom(data_config)
    g_config.test_data_config.for_test = True
    if async_load_data is not None:
        logger.warning("Deprecated: async_load_data should be used inside"
                       " Data definition")
        g_config.test_data_config.async_load_data = async_load_data

Q
qijun 已提交
1108

L
Luo Tao 已提交
1109 1110
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1111 1112 1113 1114 1115 1116 1117
def cnn_output_size(img_size, filter_size, padding, stride, caffe_mode):
    output = (2 * padding + img_size - filter_size) / float(stride)
    if caffe_mode:
        return 1 + int(math.floor(output))
    else:
        return 1 + int(math.ceil(output))

Q
qijun 已提交
1118

1119
#calcualte image_size based on output_size for de-convolution (ConvTransLayer).
L
Luo Tao 已提交
1120
#It is the reverse function of cnn_output_size
1121
def cnn_image_size(output_size, filter_size, padding, stride, caffe_mode):
L
Luo Tao 已提交
1122 1123 1124
    img_size = (output_size - 1) * stride + filter_size - 2 * padding
    if not caffe_mode:
        img_size = img_size + 1
1125 1126
    return img_size

Q
qijun 已提交
1127

L
Luo Tao 已提交
1128
def get_img_size(input_layer_name, channels):
L
Luo Tao 已提交
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
    input = g_layer_map[input_layer_name]
    img_pixels = input.size / channels
    img_size = input.width if input.width > 0 else int(img_pixels**0.5)
    img_size_y = input.height if input.height > 0 else int(img_pixels /
                                                           img_size)
    config_assert(
        img_size * img_size_y == img_pixels,
        "Input layer %s: Incorrect input image size %d * %d for input image pixels %d"
        % (input_layer_name, img_size, img_size_y, img_pixels))
    return img_size, img_size_y


def parse_bilinear(bilinear, input_layer_name, bilinear_conf):
    parse_image(bilinear, input_layer_name, bilinear_conf.image_conf)
    bilinear_conf.out_size_x = bilinear.out_size_x
    bilinear_conf.out_size_y = bilinear.out_size_y


1147
def parse_pool(pool, input_layer_name, pool_conf, ceil_mode):
Z
zhangjinchao01 已提交
1148
    pool_conf.pool_type = pool.pool_type
Q
qijun 已提交
1149 1150 1151
    config_assert(pool.pool_type in [
        'max-projection', 'avg-projection', 'cudnn-max-pool', 'cudnn-avg-pool'
    ], "pool-type %s is not in "
Z
zhangjinchao01 已提交
1152
                  "['max-projection', 'avg-projection', "
Q
qijun 已提交
1153
                  "'cudnn-max-pool', 'cudnn-avg-pool']" % pool.pool_type)
Z
zhangjinchao01 已提交
1154 1155 1156 1157 1158 1159

    pool_conf.channels = pool.channels
    pool_conf.size_x = pool.size_x
    pool_conf.stride = pool.stride

    pool_conf.size_y = default(pool.size_y, pool_conf.size_x)
Q
qijun 已提交
1160
    pool_conf.stride_y = default(pool.stride_y, pool_conf.stride)
Z
zhangjinchao01 已提交
1161

L
Luo Tao 已提交
1162
    pool_conf.img_size, pool_conf.img_size_y = \
L
Luo Tao 已提交
1163
        get_img_size(input_layer_name, pool.channels)
Z
zhangjinchao01 已提交
1164

1165
    config_assert(not pool.start, "start is deprecated in pooling.")
Z
zhangjinchao01 已提交
1166

1167
    if pool.padding is not None:
Z
zhangjinchao01 已提交
1168
        pool_conf.padding = pool.padding
1169
    pool_conf.padding_y = default(pool.padding_y, pool_conf.padding)
D
dangqingqing 已提交
1170 1171
    pool_conf.output_x = cnn_output_size(pool_conf.img_size, pool_conf.size_x,
                                         pool_conf.padding, pool_conf.stride,
1172
                                         not ceil_mode)
D
dangqingqing 已提交
1173 1174
    pool_conf.output_y = cnn_output_size(pool_conf.img_size_y, pool_conf.size_y,
                                         pool_conf.padding_y,
1175
                                         pool_conf.stride_y, not ceil_mode)
Q
qijun 已提交
1176

Z
zhangjinchao01 已提交
1177

Q
qijun 已提交
1178
def parse_spp(spp, input_layer_name, spp_conf):
L
Luo Tao 已提交
1179
    parse_image(spp, input_layer_name, spp_conf.image_conf)
Q
qijun 已提交
1180 1181
    spp_conf.pool_type = spp.pool_type
    config_assert(spp.pool_type in ['max-projection', 'avg-projection'],
Q
qijun 已提交
1182 1183
                  "pool-type %s is not in "
                  "['max-projection', 'avg-projection']" % spp.pool_type)
Q
qijun 已提交
1184
    spp_conf.pyramid_height = spp.pyramid_height
Q
qijun 已提交
1185

Q
qijun 已提交
1186

Z
zhangjinchao01 已提交
1187 1188
def parse_image(image, input_layer_name, image_conf):
    image_conf.channels = image.channels
L
Luo Tao 已提交
1189
    image_conf.img_size, image_conf.img_size_y = \
L
Luo Tao 已提交
1190
        get_img_size(input_layer_name, image_conf.channels)
Q
qijun 已提交
1191

Z
zhangjinchao01 已提交
1192 1193 1194

def parse_norm(norm, input_layer_name, norm_conf):
    norm_conf.norm_type = norm.norm_type
1195 1196 1197 1198 1199
    config_assert(
        norm.norm_type in
        ['rnorm', 'cmrnorm-projection', 'cross-channel-norm'],
        "norm-type %s is not in [rnorm, cmrnorm-projection, cross-channel-norm]"
        % norm.norm_type)
Z
zhangjinchao01 已提交
1200 1201 1202 1203 1204 1205
    norm_conf.channels = norm.channels
    norm_conf.size = norm.size
    norm_conf.scale = norm.scale
    norm_conf.pow = norm.pow
    norm_conf.blocked = norm.blocked

L
Luo Tao 已提交
1206
    norm_conf.img_size, norm_conf.img_size_y = \
L
Luo Tao 已提交
1207
        get_img_size(input_layer_name, norm.channels)
Z
zhangjinchao01 已提交
1208
    norm_conf.output_x = norm_conf.img_size
L
Luo Tao 已提交
1209
    norm_conf.output_y = norm_conf.img_size_y
Z
zhangjinchao01 已提交
1210 1211 1212
    if norm.norm_type in ['cmrnorm-projection']:
        norm_conf.scale /= norm.size
    else:
Q
qijun 已提交
1213 1214
        norm_conf.scale /= norm.size**2

1215

L
Luo Tao 已提交
1216 1217
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1218
def parse_conv(conv, input_layer_name, conv_conf, num_filters, trans=False):
Z
zhangjinchao01 已提交
1219 1220 1221 1222 1223 1224 1225 1226 1227
    conv_conf.filter_size = conv.filter_size
    conv_conf.filter_size_y = conv.filter_size_y
    conv_conf.channels = conv.channels
    conv_conf.padding = conv.padding
    conv_conf.padding_y = conv.padding_y
    conv_conf.stride = conv.stride
    conv_conf.stride_y = conv.stride_y
    conv_conf.groups = conv.groups
    conv_conf.caffe_mode = conv.caffe_mode
Q
qijun 已提交
1228

1229
    if not trans:
1230
        conv_conf.filter_channels = conv.channels / conv.groups
L
Luo Tao 已提交
1231
        conv_conf.img_size, conv_conf.img_size_y = \
L
Luo Tao 已提交
1232
            get_img_size(input_layer_name, conv.channels)
1233
        conv_conf.output_x = cnn_output_size(
Q
qijun 已提交
1234 1235
            conv_conf.img_size, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1236 1237 1238
        conv_conf.output_y = cnn_output_size(
            conv_conf.img_size_y, conv_conf.filter_size_y, conv_conf.padding_y,
            conv_conf.stride_y, conv_conf.caffe_mode)
1239
    else:
1240
        conv_conf.filter_channels = num_filters / conv.groups
L
Luo Tao 已提交
1241
        conv_conf.output_x, conv_conf.output_y = \
L
Luo Tao 已提交
1242
            get_img_size(input_layer_name, conv.channels)
1243
        conv_conf.img_size = cnn_image_size(
Q
qijun 已提交
1244 1245
            conv_conf.output_x, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1246
        conv_conf.img_size_y = cnn_image_size(
L
Luo Tao 已提交
1247 1248
            conv_conf.output_y, conv_conf.filter_size_y, conv_conf.padding_y,
            conv_conf.stride_y, conv_conf.caffe_mode)
Q
qijun 已提交
1249

1250

Z
zhangjinchao01 已提交
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263
def parse_block_expand(block_expand, input_layer_name, block_expand_conf):
    block_expand_conf.channels = block_expand.channels
    block_expand_conf.stride_x = block_expand.stride_x
    block_expand_conf.stride_y = block_expand.stride_y
    block_expand_conf.padding_x = block_expand.padding_x
    block_expand_conf.padding_y = block_expand.padding_y
    block_expand_conf.block_x = block_expand.block_x
    block_expand_conf.block_y = block_expand.block_y
    block_expand_conf.img_size_x = block_expand.img_size_x
    block_expand_conf.img_size_y = block_expand.img_size_y
    if block_expand_conf.img_size_x == 0:
        block_expand_conf.output_x = 0
    else:
1264
        block_expand_conf.output_x = cnn_output_size(
1265
            block_expand.img_size_x, block_expand.block_x,
1266
            block_expand.padding_x, block_expand.stride_x, False)
Z
zhangjinchao01 已提交
1267 1268

    if block_expand_conf.img_size_y == 0:
1269
        block_expand_conf.output_y = 0
Z
zhangjinchao01 已提交
1270
    else:
1271
        block_expand_conf.output_y = cnn_output_size(
1272
            block_expand.img_size_y, block_expand.block_y,
1273
            block_expand.padding_y, block_expand.stride_y, False)
Z
zhangjinchao01 已提交
1274

Q
qijun 已提交
1275

1276
def parse_maxout(maxout, input_layer_name, maxout_conf):
L
Luo Tao 已提交
1277
    parse_image(maxout, input_layer_name, maxout_conf.image_conf)
1278
    maxout_conf.groups = maxout.groups
1279

Q
qijun 已提交
1280

Z
zhangjinchao01 已提交
1281 1282
# Define an evaluator
@config_func
Y
yangyaming 已提交
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
def Evaluator(name,
              type,
              inputs,
              chunk_scheme=None,
              num_chunk_types=None,
              classification_threshold=None,
              positive_label=None,
              dict_file=None,
              result_file=None,
              num_results=None,
              top_k=None,
              delimited=None,
              excluded_chunk_types=None,
              overlap_threshold=None,
              background_id=None,
              evaluate_difficult=None,
              ap_type=None):
Z
zhangjinchao01 已提交
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
    evaluator = g_config.model_config.evaluators.add()
    evaluator.type = type
    evaluator.name = MakeLayerNameInSubmodel(name)
    if type_of(inputs) == str:
        inputs = [inputs]

    evaluator.input_layers.extend(
        [MakeLayerNameInSubmodel(name) for name in inputs])

    if chunk_scheme is not None:
        evaluator.chunk_scheme = chunk_scheme
        evaluator.num_chunk_types = num_chunk_types
    g_current_submodel.evaluator_names.append(evaluator.name)

1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
    if classification_threshold is not None:
        evaluator.classification_threshold = classification_threshold
    if positive_label is not None:
        evaluator.positive_label = positive_label
    if dict_file is not None:
        evaluator.dict_file = dict_file

    if result_file is not None:
        evaluator.result_file = result_file
    if num_results is not None:
        evaluator.num_results = num_results
L
Liang Zhao 已提交
1325 1326
    if top_k is not None:
        evaluator.top_k = top_k
1327 1328
    if delimited is not None:
        evaluator.delimited = delimited
Z
zhangjinchao01 已提交
1329

1330 1331 1332
    if excluded_chunk_types:
        evaluator.excluded_chunk_types.extend(excluded_chunk_types)

Y
yangyaming 已提交
1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344
    if overlap_threshold is not None:
        evaluator.overlap_threshold = overlap_threshold

    if background_id is not None:
        evaluator.background_id = background_id

    if evaluate_difficult is not None:
        evaluator.evaluate_difficult = evaluate_difficult

    if ap_type is not None:
        evaluator.ap_type = ap_type

Q
qijun 已提交
1345

Z
zhangjinchao01 已提交
1346 1347 1348 1349 1350
class LayerBase(object):
    def __init__(
            self,
            name,
            type,
Q
qijun 已提交
1351
            size,  # size can be 0. In this case, subclass should set it.
Z
zhangjinchao01 已提交
1352 1353 1354 1355
            inputs,
            device=None,
            active_type="",
            drop_rate=0.,
C
caoying03 已提交
1356 1357
            coeff=None,
            error_clipping_threshold=None):
Z
zhangjinchao01 已提交
1358
        config_assert('@' not in name,
Q
qijun 已提交
1359
                      "layer name: %s contain special character @" % name)
Z
zhangjinchao01 已提交
1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
        global g_current_submodel
        name = MakeLayerNameInSubmodel(name)

        config_assert(name not in g_layer_map,
                      'Duplicated layer name: %s' % name)

        self.inputs = copy.deepcopy(inputs)
        self.operators = []

        if self.inputs is None:
            self.inputs = []
        elif type_of(self.inputs) != list:
            self.inputs = [self.inputs]

        self.config = g_config.model_config.layers.add()
1375
        assert isinstance(self.config, LayerConfig)
Z
zhangjinchao01 已提交
1376 1377 1378
        self.config.name = name
        self.config.type = type
        self.config.active_type = active_type
1379 1380
        if coeff is not None:
            self.config.coeff = float(coeff)
Z
zhangjinchao01 已提交
1381 1382 1383 1384 1385 1386 1387
        if size != 0:
            self.config.size = size
        if drop_rate != 0:
            self.config.drop_rate = drop_rate

        if device is not None:
            self.config.device = device
1388
        elif g_default_device is not None:
Z
zhangjinchao01 已提交
1389 1390
            self.config.device = g_default_device

C
caoying03 已提交
1391 1392 1393
        if error_clipping_threshold is not None:
            self.config.error_clipping_threshold = error_clipping_threshold

Z
zhangjinchao01 已提交
1394 1395 1396 1397 1398 1399 1400
        for input_index in xrange(len(self.inputs)):
            input = self.inputs[input_index]
            input_config = None
            input_layer_name = ''
            if type_of(input) == str:
                input_layer_name = input
                input_config = Input(
Q
qijun 已提交
1401 1402
                    input_layer_name=input,
                    parameter_name=gen_parameter_name(name, input_index))
Z
zhangjinchao01 已提交
1403 1404 1405 1406 1407 1408 1409 1410
                input_layer_name = input_config.input_layer_name
            elif isinstance(input, Input):
                input_layer_name = input.input_layer_name
                input_config = input
                if input_config.parameter_name is None:
                    input_config.parameter_name = \
                        gen_parameter_name(name, input_index)
            elif isinstance(input, Operator):
Q
qijun 已提交
1411
                self.operators.append(input)
Z
zhangjinchao01 已提交
1412 1413 1414 1415
                input.operator_conf.input_indices.append(input_index)
                input_config = Input(input.input_layer_names[0])
                input_layer_name = input_config.input_layer_name
            else:
Q
qijun 已提交
1416
                raise ValueError('Wrong type for inputs: %s' % type_of(input))
Z
zhangjinchao01 已提交
1417
            config_assert(input_layer_name in g_layer_map,
Q
qijun 已提交
1418 1419
                          "Unknown input layer '%s' for layer %s" %
                          (input_layer_name, name))
Z
zhangjinchao01 已提交
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
            self.inputs[input_index] = input_config
            layer_input = self.config.inputs.add()
            layer_input.input_layer_name = input_config.input_layer_name
            if input_config.input_layer_argument is not None:
                layer_input.input_layer_argument = \
                    input_config.input_layer_argument

        g_layer_map[name] = self.config

        g_current_submodel.layer_names.append(self.config.name)

    def get_input_layer(self, input_index):
        return g_layer_map[self.config.inputs[input_index].input_layer_name]

    # will return the bias created if not *for_self*
    def create_bias_parameter(
            self,
Q
qijun 已提交
1437
            bias,  # True/False or BiasCfg
Z
zhangjinchao01 已提交
1438
            size,
Q
qijun 已提交
1439 1440 1441
            dims=None,
            for_self=True,  # whether create bias for layer self
    ):
Z
zhangjinchao01 已提交
1442 1443 1444 1445 1446 1447

        if size == 0:
            return
        if dims is None:
            dims = [1, size]

Q
qijun 已提交
1448 1449 1450
        config_assert(
            type_of(bias) == bool or type_of(bias) == Bias,
            'Incorrect type for bias: %s' % type_of(bias))
Z
zhangjinchao01 已提交
1451 1452 1453 1454 1455 1456 1457 1458 1459

        if type_of(bias) == bool:
            if bias:
                bias = Bias()

        if type_of(bias) == Bias:
            if bias.parameter_name is None:
                bias.parameter_name = gen_bias_parameter_name(self.config.name)
            if bias.parameter_name not in g_parameter_map:
1460 1461
                assert isinstance(self.config, LayerConfig)

Z
zhangjinchao01 已提交
1462 1463 1464
                Parameter(
                    bias.parameter_name,
                    size,
Q
qijun 已提交
1465 1466
                    self.config.device
                    if self.config.HasField('device') else None,
Z
zhangjinchao01 已提交
1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
                    dims,
                    bias.learning_rate,
                    bias.momentum,
                    decay_rate=bias.decay_rate,
                    decay_rate_l1=bias.decay_rate_l1,
                    initial_mean=bias.initial_mean,
                    initial_std=bias.initial_std,
                    initial_strategy=bias.initial_strategy,
                    initial_smart=bias.initial_smart,
                    num_batches_regularization=bias.num_batches_regularization,
                    sparse_remote_update=bias.sparse_remote_update,
Q
qijun 已提交
1478 1479
                    gradient_clipping_threshold=bias.
                    gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1480
                    is_static=bias.is_static,
X
xuwei06 已提交
1481 1482
                    is_shared=bias.is_shared,
                    initializer=bias.initializer)
Z
zhangjinchao01 已提交
1483 1484 1485 1486 1487
            if for_self:
                self.config.bias_parameter_name = bias.parameter_name
            else:
                return bias.parameter_name

Q
qijun 已提交
1488 1489 1490 1491 1492 1493
    def create_input_parameter(self,
                               input_index,
                               size,
                               dims=None,
                               sparse=None,
                               format=None):
Z
zhangjinchao01 已提交
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
        if dims is None:
            # TODO(yuyang18): print warning and callstack here!
            dims = list()

        if size == 0:
            return

        input_config = self.inputs[input_index]

        self.config.inputs[input_index].input_parameter_name = \
            input_config.parameter_name

        if input_config.parameter_name in g_parameter_map:
            para = g_parameter_map[input_config.parameter_name]
Q
qijun 已提交
1508 1509
            config_assert(size == para.size, (
                'Shared parameter "%s" does not ' + 'have same size: %s vs. %s')
Z
zhangjinchao01 已提交
1510 1511
                          % (input_config.parameter_name, para.size, size))

Q
qijun 已提交
1512 1513
            config_assert(dims == para.dims, (
                'Shared parameter "%s" does not ' + 'have same dims: %s vs. %s')
Z
zhangjinchao01 已提交
1514 1515 1516 1517 1518 1519
                          % (input_config.parameter_name, para.dims, dims))
            return

        Parameter(
            input_config.parameter_name,
            size,
1520
            self.config.device if self.config.HasField("device") else None,
Z
zhangjinchao01 已提交
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
            dims,
            input_config.learning_rate,
            input_config.momentum,
            decay_rate=input_config.decay_rate,
            decay_rate_l1=input_config.decay_rate_l1,
            initial_mean=input_config.initial_mean,
            initial_std=input_config.initial_std,
            initial_strategy=input_config.initial_strategy,
            initial_smart=input_config.initial_smart,
            num_batches_regularization=input_config.num_batches_regularization,
            sparse_remote_update=input_config.sparse_remote_update,
            sparse_update=input_config.sparse_update,
Q
qijun 已提交
1533 1534
            gradient_clipping_threshold=input_config.
            gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1535 1536 1537 1538
            sparse=sparse,
            format=format,
            is_static=input_config.is_static,
            is_shared=input_config.is_shared,
X
xuwei06 已提交
1539 1540
            update_hooks=input_config.update_hooks,
            initializer=input_config.initializer)
Z
zhangjinchao01 已提交
1541 1542 1543 1544 1545 1546 1547 1548 1549

    def set_layer_size(self, size):
        if self.config.size == 0:
            self.config.size = size
        else:
            config_assert(self.config.size == size,
                          'Different inputs result in' +
                          'different layer size at layer %s' % self.config.name)

L
Luo Tao 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
    def set_layer_height_width(self, height, width):
        self.config.height = height
        self.config.width = width

    def set_cnn_layer(self,
                      input_layer_name,
                      height,
                      width,
                      channels,
                      is_print=True):
        size = height * width * channels
        self.set_layer_size(size)
        self.set_layer_height_width(height, width)
        if is_print:
            print("output for %s: c = %d, h = %d, w = %d, size = %d" %
                  (input_layer_name, channels, height, width, size))

Q
qijun 已提交
1567

Z
zhangjinchao01 已提交
1568 1569
@config_layer('multi_class_cross_entropy_with_selfnorm')
class MultiClassCrossEntropySelfNormCostLayer(LayerBase):
Q
qijun 已提交
1570 1571 1572
    def __init__(self, name, inputs, softmax_selfnorm_alpha=0.1, **xargs):
        super(MultiClassCrossEntropySelfNormCostLayer, self).__init__(
            name, 'multi_class_cross_entropy_with_selfnorm', 0, inputs, **xargs)
Z
zhangjinchao01 已提交
1573 1574
        self.config.softmax_selfnorm_alpha = softmax_selfnorm_alpha

Q
qijun 已提交
1575

Z
zhangjinchao01 已提交
1576 1577
@config_layer('fc')
class FCLayer(LayerBase):
L
lianxiaochen 已提交
1578 1579 1580 1581 1582 1583 1584
    def __init__(self,
                 name,
                 size,
                 inputs,
                 bias=True,
                 error_clipping_threshold=None,
                 **xargs):
Z
zhangjinchao01 已提交
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594
        super(FCLayer, self).__init__(name, 'fc', size, inputs=inputs, **xargs)
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            psize = self.config.size * input_layer.size
            dims = [input_layer.size, self.config.size]
            format = self.inputs[input_index].format
            sparse = format == "csr" or format == "csc"

            if sparse:
                psize = self.inputs[input_index].nnz
1595 1596
            else:
                sparse = None
Z
zhangjinchao01 已提交
1597

Q
qijun 已提交
1598 1599
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1600
        self.create_bias_parameter(bias, self.config.size)
L
lianxiaochen 已提交
1601 1602
        if error_clipping_threshold is not None:
            self.config.error_clipping_threshold = error_clipping_threshold
Z
zhangjinchao01 已提交
1603

Q
qijun 已提交
1604

Z
zhangjinchao01 已提交
1605 1606
@config_layer('selective_fc')
class SelectiveFCLayer(LayerBase):
Q
qijun 已提交
1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
    def __init__(self,
                 name,
                 size,
                 inputs,
                 bias=True,
                 selective_fc_pass_generation=False,
                 has_selected_colums=True,
                 selective_fc_full_mul_ratio=0.02,
                 selective_fc_parallel_plain_mul_thread_num=None,
                 **xargs):
Z
zhangjinchao01 已提交
1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
        super(SelectiveFCLayer, self).__init__(
            name, 'selective_fc', size, inputs=inputs, **xargs)
        # user MUST know if selctive fc is used in training,
        # parameter matrices saved by this layer are automatically transposed,
        # BUT bias is not.

        # if selective_fc is used only in testing mode, and parameters for
        # this layer are trained by fully connected layers,
        # then TranposedFullMatrixProjectin MUST be used in training
        # to avoid manual transpose in testing.

        self.config.selective_fc_pass_generation = selective_fc_pass_generation
        self.config.has_selected_colums = has_selected_colums
        self.config.selective_fc_full_mul_ratio = selective_fc_full_mul_ratio
        if selective_fc_parallel_plain_mul_thread_num is not None:
            self.config.selective_fc_parallel_plain_mul_thread_num = selective_fc_parallel_plain_mul_thread_num

        input_num = len(self.inputs)
        if has_selected_colums:
            config_assert(input_num >= 2,
Q
qijun 已提交
1637 1638
                          ("if indices of selected columns are not specified, "
                           "selective_fc Layer has at least two inputs"))
Z
zhangjinchao01 已提交
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
            input_num -= 1

        for input_index in xrange(input_num):
            input_layer = self.get_input_layer(input_index)
            psize = self.config.size * input_layer.size
            dims = [input_layer.size, self.config.size]
            dims = dims[::-1]  # transpose the parameter
            format = self.inputs[input_index].format
            sparse = format == "csr" or format == "csc"
            if sparse:
                psize = self.inputs[input_index].nnz

Q
qijun 已提交
1651 1652
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1653 1654
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1655

1656 1657
@config_layer('print')
class PrintLayer(LayerBase):
1658
    def __init__(self, name, inputs, format=None):
1659
        super(PrintLayer, self).__init__(name, 'print', 0, inputs)
1660 1661 1662 1663 1664 1665
        if format is None:
            format = "\n".join([
                "layer=" + input.input_layer_name + " %s"
                for input in self.inputs
            ])
        self.config.user_arg = format
1666

Q
qijun 已提交
1667

Y
yuan 已提交
1668 1669
@config_layer('priorbox')
class PriorBoxLayer(LayerBase):
G
gaoyuan 已提交
1670 1671
    def __init__(self, name, inputs, size, min_size, max_size, aspect_ratio,
                 variance):
Y
yuan 已提交
1672
        super(PriorBoxLayer, self).__init__(name, 'priorbox', 0, inputs)
G
gaoyuan 已提交
1673
        config_assert(len(inputs) == 2, 'PriorBoxLayer must have 2 inputs')
G
gaoyuan 已提交
1674 1675 1676 1677 1678 1679 1680
        input_layer = self.get_input_layer(1)
        config_assert(
            input_layer.type == 'data',
            'Expecting the second input layer of an priorbox layer to be '
            'a data layer')
        config_assert(input_layer.width > 0, 'The data layer must set width')
        config_assert(input_layer.height > 0, 'The data layer must set height')
G
gaoyuan 已提交
1681
        config_assert(len(variance) == 4, 'The variance must have 4 inputs')
Y
yuan 已提交
1682 1683 1684 1685 1686 1687
        self.config.inputs[0].priorbox_conf.min_size.extend(min_size)
        self.config.inputs[0].priorbox_conf.max_size.extend(max_size)
        self.config.inputs[0].priorbox_conf.aspect_ratio.extend(aspect_ratio)
        self.config.inputs[0].priorbox_conf.variance.extend(variance)
        self.config.size = size

Q
qijun 已提交
1688

1689 1690 1691
@config_layer('multibox_loss')
class MultiBoxLossLayer(LayerBase):
    def __init__(self, name, inputs, input_num, num_classes, overlap_threshold,
1692
                 neg_pos_ratio, neg_overlap, background_id, **xargs):
1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713
        super(MultiBoxLossLayer, self).__init__(name, 'multibox_loss', 0,
                                                inputs)
        config_assert(
            len(inputs) == (input_num * 2 + 2),
            'MultiBoxLossLayer does not have enough inputs')
        config_assert(num_classes > background_id,
                      'Classes number must greater than background ID')
        self.config.inputs[0].multibox_loss_conf.num_classes = num_classes
        self.config.inputs[
            0].multibox_loss_conf.overlap_threshold = overlap_threshold
        self.config.inputs[0].multibox_loss_conf.neg_pos_ratio = neg_pos_ratio
        self.config.inputs[0].multibox_loss_conf.neg_overlap = neg_overlap
        self.config.inputs[0].multibox_loss_conf.background_id = background_id
        self.config.inputs[0].multibox_loss_conf.input_num = input_num
        self.config.size = 1


@config_layer('detection_output')
class DetectionOutputLayer(LayerBase):
    def __init__(self, name, inputs, size, input_num, num_classes,
                 nms_threshold, nms_top_k, keep_top_k, confidence_threshold,
1714
                 background_id, **xargs):
1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
        super(DetectionOutputLayer, self).__init__(name, 'detection_output', 0,
                                                   inputs)
        config_assert(
            len(inputs) == (input_num * 2 + 1),
            'DetectionOutputLayer does not have enough inputs')
        config_assert(num_classes > background_id,
                      'Classes number must greater than background ID')
        self.config.inputs[0].detection_output_conf.num_classes = num_classes
        self.config.inputs[
            0].detection_output_conf.nms_threshold = nms_threshold
        self.config.inputs[0].detection_output_conf.nms_top_k = nms_top_k
        self.config.inputs[0].detection_output_conf.keep_top_k = keep_top_k
        self.config.inputs[
            0].detection_output_conf.confidence_threshold = confidence_threshold
        self.config.inputs[
            0].detection_output_conf.background_id = background_id
        self.config.inputs[0].detection_output_conf.input_num = input_num
        self.config.size = size


Z
zhangjinchao01 已提交
1735 1736
@config_layer('data')
class DataLayer(LayerBase):
L
Luo Tao 已提交
1737
    def __init__(self, name, size, height=None, width=None, device=None):
Q
qijun 已提交
1738 1739
        super(DataLayer, self).__init__(
            name, 'data', size, inputs=[], device=device)
L
Luo Tao 已提交
1740 1741
        if height and width:
            self.set_layer_height_width(height, width)
Q
qijun 已提交
1742

Z
zhangjinchao01 已提交
1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769

'''
DataNormLayer: A layer for data normalization
Input: One and only one input layer is accepted. The input layer must
       be DataLayer with dense data type
Output: The normalization of the input data

Reference:
    LA Shalabi, Z Shaaban, B Kasasbeh. Data mining: A preprocessing engine

Example:
    Layer(
        name = "norm_input_layer",
        type = "data_norm",
        inputs = [Input("input_layer",
                        parameter_name = "_slot0.stats")],
        data_norm_strategy = "z-score",
    )

Note:
  (1) The parameter has been calculated in the preprocessing stage,
      and should be initialized by --init_model_path when training.
  (2) Three data normalization methoeds are considered
          z-score: y = (x-mean)/std
          min-max: y = (x-min)/(max-min)
          decimal-scaling: y = x/10^j, where j is the smallest integer such that max(|y|)<1
'''
Q
qijun 已提交
1770 1771


Z
zhangjinchao01 已提交
1772 1773
@config_layer('data_norm')
class DataNormLayer(LayerBase):
Q
qijun 已提交
1774
    def __init__(self, name, inputs, data_norm_strategy="z-score", device=None):
Z
zhangjinchao01 已提交
1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785
        super(DataNormLayer, self).__init__(
            name, 'data_norm', 0, inputs=inputs, device=device)
        self.config.data_norm_strategy = data_norm_strategy
        config_assert(len(inputs) == 1, 'DataNormLayer must have 1 input')
        input_layer = self.get_input_layer(0)
        self.set_layer_size(input_layer.size)
        para_size = 5 * input_layer.size
        para_dims = [5, input_layer.size]
        self.inputs[0].is_static = True
        self.create_input_parameter(0, para_size, para_dims)

Q
qijun 已提交
1786

Z
zhangjinchao01 已提交
1787 1788 1789
@config_layer('prelu')
class ParameterReluLayer(LayerBase):
    layer_type = 'prelu'
Q
qijun 已提交
1790 1791

    def __init__(self, name, inputs, partial_sum=1, **args):
Z
zhangjinchao01 已提交
1792 1793 1794
        super(ParameterReluLayer, self).__init__(
            name, self.layer_type, 0, inputs=inputs, **args)
        input_layer = self.get_input_layer(0)
1795 1796 1797
        config_assert(len(self.inputs) == 1, "prelu layer has only one input.")
        config_assert(input_layer.size % partial_sum == 0,
                      "a wrong setting for partial_sum")
Z
zhangjinchao01 已提交
1798 1799 1800
        self.set_layer_size(input_layer.size)
        self.create_input_parameter(0, input_layer.size / partial_sum)

Q
qijun 已提交
1801

Z
zhangjinchao01 已提交
1802 1803 1804
@config_layer('conv')
class ConvLayerBase(LayerBase):
    layer_type = 'conv'
Q
qijun 已提交
1805 1806 1807 1808 1809 1810 1811 1812

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
Z
zhangjinchao01 已提交
1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
        super(ConvLayerBase, self).__init__(
            name, self.layer_type, 0, inputs=inputs, **xargs)

        if num_filters is not None:
            self.config.num_filters = num_filters

        use_gpu = int(g_command_config_args.get("use_gpu", 0))
        parallel_nn = int(g_command_config_args.get("parallel_nn", 0))

        # Automatically select cudnn_type for GPU and exconv for CPU
        # if set type=conv, but still reserve the way user specify
        # exconv or cudnn_conv manually.
        if self.layer_type == "cudnn_conv":
            config_assert(use_gpu, "cudnn_conv only support GPU")

        if (use_gpu == 1 and self.layer_type != "exconv" and
Q
qijun 已提交
1829
            (parallel_nn == 0 or self.config.device > -1)):
Z
zhangjinchao01 已提交
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
            self.layer_type = "cudnn_conv"
        else:
            self.layer_type = "exconv"
        # need to specify layer in config
        self.config.type = self.layer_type

        if shared_biases is not None:
            self.config.shared_biases = shared_biases

        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            conv_conf = self.config.inputs[input_index].conv_conf
L
Luo Tao 已提交
1842 1843
            parse_conv(self.inputs[input_index].conv, input_layer.name,
                       conv_conf, num_filters)
Z
zhangjinchao01 已提交
1844 1845
            psize = self.calc_parameter_size(conv_conf)
            self.create_input_parameter(input_index, psize)
L
Luo Tao 已提交
1846 1847
            self.set_cnn_layer(name, conv_conf.output_y, conv_conf.output_x,
                               self.config.num_filters)
Z
zhangjinchao01 已提交
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857

        psize = self.config.size
        if shared_biases:
            psize = self.config.num_filters
        self.create_bias_parameter(bias, psize, [psize, 1])

    def calc_parameter_size(self, conv_conf):
        return self.config.num_filters * conv_conf.filter_channels \
                    * (conv_conf.filter_size * conv_conf.filter_size_y)

Q
qijun 已提交
1858

Z
zhangjinchao01 已提交
1859 1860 1861 1862
@config_layer('exconv')
class ConvLayer(ConvLayerBase):
    layer_type = 'exconv'

Q
qijun 已提交
1863

Z
zhangjinchao01 已提交
1864 1865 1866 1867
@config_layer('cudnn_conv')
class ConvLayer(ConvLayerBase):
    layer_type = 'cudnn_conv'

1868 1869 1870 1871

@config_layer('convt')
class ConvTransLayerBase(LayerBase):
    layer_type = 'convt'
Q
qijun 已提交
1872 1873 1874 1875 1876 1877 1878 1879

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
1880
        super(ConvTransLayerBase, self).__init__(
1881 1882 1883 1884 1885 1886 1887 1888
            name, self.layer_type, 0, inputs=inputs, **xargs)

        if num_filters is not None:
            self.config.num_filters = num_filters

        use_gpu = int(g_command_config_args.get("use_gpu", 0))
        parallel_nn = int(g_command_config_args.get("parallel_nn", 0))

1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
        # Automatically select cudnn_type for GPU and exconvt for CPU
        # if set type=exconvt, but still reserve the way user specify
        # exconvt or cudnn_convt manually.
        if self.layer_type == "cudnn_convt":
            config_assert(use_gpu, "cudnn_convt only support GPU")

        if (use_gpu == 1 and self.layer_type != "exconvt" and
            (parallel_nn == 0 or self.config.device > -1)):
            self.layer_type = "cudnn_convt"
        else:
            self.layer_type = "exconvt"
1900 1901 1902 1903 1904 1905 1906 1907
        # need to specify layer in config
        self.config.type = self.layer_type

        if shared_biases is not None:
            self.config.shared_biases = shared_biases

        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
1908
            parse_conv(
1909 1910
                self.inputs[input_index].conv,
                input_layer.name,
1911
                self.config.inputs[input_index].conv_conf,
1912
                num_filters,
1913
                trans=True)
1914 1915 1916
            conv_conf = self.config.inputs[input_index].conv_conf
            psize = self.calc_parameter_size(conv_conf)
            self.create_input_parameter(input_index, psize)
1917 1918
            self.set_cnn_layer(name, conv_conf.img_size_y, conv_conf.img_size,
                               self.config.num_filters)
1919 1920 1921 1922 1923 1924 1925

        psize = self.config.size
        if shared_biases:
            psize = self.config.num_filters
        self.create_bias_parameter(bias, psize, [psize, 1])

    def calc_parameter_size(self, conv_conf):
1926
        return conv_conf.channels * conv_conf.filter_channels \
1927 1928
                    * (conv_conf.filter_size * conv_conf.filter_size_y)

Q
qijun 已提交
1929

1930 1931 1932 1933
@config_layer('exconvt')
class ConvTransLayer(ConvTransLayerBase):
    layer_type = 'exconvt'

Q
qijun 已提交
1934

1935 1936 1937 1938 1939
@config_layer('cudnn_convt')
class ConvTransLayer(ConvTransLayerBase):
    layer_type = 'cudnn_convt'


Z
zhangjinchao01 已提交
1940 1941
@config_layer('norm')
class NormLayer(LayerBase):
1942 1943
    def __init__(self, name, inputs, **xargs):
        super(NormLayer, self).__init__(name, 'norm', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
1944 1945 1946
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            norm_conf = self.config.inputs[input_index].norm_conf
L
Luo Tao 已提交
1947 1948 1949 1950
            parse_norm(self.inputs[input_index].norm, input_layer.name,
                       norm_conf)
            self.set_cnn_layer(name, norm_conf.output_y, norm_conf.output_x,
                               norm_conf.channels, False)
1951 1952 1953
            if norm_conf.norm_type == "cross-channel-norm":
                self.create_input_parameter(0, norm_conf.channels,
                                            [norm_conf.channels, 1])
Q
qijun 已提交
1954

Z
zhangjinchao01 已提交
1955 1956 1957

@config_layer('pool')
class PoolLayer(LayerBase):
1958 1959
    def __init__(self, name, inputs, ceil_mode=True, **xargs):
        super(PoolLayer, self).__init__(name, 'pool', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
1960 1961 1962
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            pool_conf = self.config.inputs[input_index].pool_conf
L
Luo Tao 已提交
1963
            parse_pool(self.inputs[input_index].pool, input_layer.name,
1964
                       pool_conf, ceil_mode)
L
Luo Tao 已提交
1965 1966
            self.set_cnn_layer(name, pool_conf.output_y, pool_conf.output_x,
                               pool_conf.channels)
Q
qijun 已提交
1967

Z
zhangjinchao01 已提交
1968

Q
qijun 已提交
1969 1970
@config_layer('spp')
class SpatialPyramidPoolLayer(LayerBase):
1971
    def __init__(self, name, inputs, **xargs):
Q
qijun 已提交
1972
        super(SpatialPyramidPoolLayer, self).__init__(
1973
            name, 'spp', 0, inputs=inputs, **xargs)
Q
qijun 已提交
1974 1975 1976
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            spp_conf = self.config.inputs[input_index].spp_conf
L
Luo Tao 已提交
1977 1978 1979
            parse_spp(self.inputs[input_index].spp, input_layer.name, spp_conf)
            output_x = (pow(4, spp_conf.pyramid_height) - 1) / (4 - 1)
            self.set_cnn_layer(name, 1, output_x, spp_conf.image_conf.channels)
Q
qijun 已提交
1980

Q
qijun 已提交
1981

D
dangqingqing 已提交
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
@config_layer('pad')
class PadLayer(LayerBase):
    def __init__(self, name, inputs, **xargs):
        super(PadLayer, self).__init__(name, 'pad', 0, inputs=inputs, **xargs)
        pad = self.inputs[0].pad
        self.config.inputs[0].pad_conf.pad_c.extend(pad.pad_c)
        self.config.inputs[0].pad_conf.pad_h.extend(pad.pad_h)
        self.config.inputs[0].pad_conf.pad_w.extend(pad.pad_w)

        input_layer = self.get_input_layer(0)
        image_conf = self.config.inputs[0].pad_conf.image_conf
        parse_image(pad, input_layer.name, image_conf)
        out_ch = pad.channels + pad.pad_c[0] + pad.pad_c[1]
        out_h = image_conf.img_size_y + pad.pad_h[0] + pad.pad_h[1]
        out_w = image_conf.img_size + pad.pad_w[0] + pad.pad_w[1]
        self.set_cnn_layer(name, out_h, out_w, out_ch)
        self.config.size = out_ch * out_h * out_w


2001 2002
@config_layer('crop')
class CropLayer(LayerBase):
2003
    def __init__(self, name, inputs, axis, offset, shape, **xargs):
2004
        super(CropLayer, self).__init__(name, 'crop', 0, inputs=inputs, **xargs)
2005 2006 2007
        self.config.axis = axis
        self.config.offset.extend(offset)
        self.config.shape.extend(shape)
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

        # get channel, width and height from input_0 layer
        input_layer = self.get_input_layer(0)
        image_conf = self.config.inputs[0].image_conf
        image_conf.img_size = input_layer.width
        image_conf.img_size_y = input_layer.height
        image_conf.channels = input_layer.size / (input_layer.width *
                                                  input_layer.height)


Z
zhangjinchao01 已提交
2018 2019 2020
@config_layer('batch_norm')
class BatchNormLayer(LayerBase):
    layer_type = 'batch_norm'
Q
qijun 已提交
2021 2022 2023 2024 2025 2026 2027 2028 2029

    def __init__(self,
                 name,
                 inputs,
                 bias=True,
                 use_global_stats=True,
                 moving_average_fraction=0.9,
                 batch_norm_type=None,
                 **xargs):
Z
zhangjinchao01 已提交
2030 2031 2032 2033
        if inputs is None:
            inputs = []
        elif not isinstance(inputs, list):
            inputs = [inputs]
Q
qijun 已提交
2034 2035
        config_assert(
            len(inputs) == 1, "BatchNormLayer must have one and only one input")
Z
zhangjinchao01 已提交
2036 2037 2038 2039 2040 2041 2042 2043
        # Create Input for moving mean and std,
        # in batch normalization layer.
        # These paras no need to update, so set is_static is true.
        # If not use is_static, even set learning_rate = 0, decay_rate = 0,
        # these paras will change if set average_window in configure.
        use_gpu = bool(int(g_command_config_args.get("use_gpu", 0)))
        is_shared = True if not use_gpu else False
        for i in xrange(2):
Q
qijun 已提交
2044 2045 2046 2047 2048 2049
            inputs.append(
                Input(
                    inputs[0].input_layer_name,
                    initial_std=0.0,
                    initial_mean=0.0,
                    is_static=True,
2050
                    is_shared=is_shared,
D
dangqingqing 已提交
2051
                    make_layer_name_in_submodel=False, ))
Z
zhangjinchao01 已提交
2052 2053 2054 2055 2056 2057

        parallel_nn = bool(int(g_command_config_args.get("parallel_nn", 0)))
        cudnn_version = int(g_command_config_args.get("cudnn_version", 0))
        # Automatically select cudnn_batch_norm for GPU and batch_norm for CPU.
        # Also based on cudnn version.
        use_cudnn = use_gpu and batch_norm_type != "batch_norm" and \
2058
                ((not parallel_nn) or self.config.device > -1)
Z
zhangjinchao01 已提交
2059
        self.layer_type = "cudnn_batch_norm" if use_cudnn else "batch_norm"
Q
qijun 已提交
2060
        super(BatchNormLayer, self).__init__(
X
xuwei06 已提交
2061
            name, self.layer_type, 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2062 2063 2064 2065 2066 2067

        if use_global_stats is not None:
            self.config.use_global_stats = use_global_stats
        if moving_average_fraction is not None:
            self.config.moving_average_fraction = moving_average_fraction

Q
qijun 已提交
2068
        input_layer = self.get_input_layer(0)
Z
zhangjinchao01 已提交
2069
        image_conf = self.config.inputs[0].image_conf
L
Luo Tao 已提交
2070
        parse_image(self.inputs[0].image, input_layer.name, image_conf)
2071

2072 2073
        # Only pass the width and height of input to batch_norm layer
        # when either of it is non-zero.
2074 2075
        if input_layer.width != 0 or input_layer.height != 0:
            self.set_cnn_layer(name, image_conf.img_size_y, image_conf.img_size,
D
dangqingqing 已提交
2076
                               image_conf.channels, False)
2077 2078
        else:
            self.set_layer_size(input_layer.size)
Z
zhangjinchao01 已提交
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090

        psize = self.calc_parameter_size(image_conf)
        dims = [1, psize]
        self.create_input_parameter(0, psize)
        self.create_input_parameter(1, psize, dims)
        self.create_input_parameter(2, psize, dims)

        self.create_bias_parameter(bias, psize)

    def calc_parameter_size(self, image_conf):
        return image_conf.channels

Q
qijun 已提交
2091

Z
zhangjinchao01 已提交
2092 2093
@config_layer('trans')
class TransLayer(LayerBase):
2094
    def __init__(self, name, inputs, **xargs):
Q
qijun 已提交
2095
        super(TransLayer, self).__init__(
2096
            name, 'trans', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2097 2098 2099
        config_assert(
            len(self.inputs) == 1,
            'TransLayer must have one and only one input')
Z
zhangjinchao01 已提交
2100 2101
        self.set_layer_size(self.get_input_layer(0).size)

Q
qijun 已提交
2102

Z
zhangjinchao01 已提交
2103 2104
@config_layer('resize')
class ResizeLayer(LayerBase):
2105
    def __init__(self, name, size, inputs, **xargs):
Q
qijun 已提交
2106
        super(ResizeLayer, self).__init__(
2107
            name, 'resize', size=size, inputs=inputs, **xargs)
Q
qijun 已提交
2108 2109 2110 2111
        config_assert(
            len(self.inputs) == 1,
            'ResizeLayer must have one and only one input')

Z
zhangjinchao01 已提交
2112

2113 2114
@config_layer('rotate')
class RotateLayer(LayerBase):
H
Haonan 已提交
2115
    def __init__(self, name, inputs, height, width, device=None):
2116 2117 2118 2119 2120
        super(RotateLayer, self).__init__(
            name, 'rotate', 0, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 1,
            'RotateLayer must have one and only one input')
H
Haonan 已提交
2121
        self.set_layer_height_width(height, width)
2122 2123 2124
        self.set_layer_size(self.get_input_layer(0).size)


Z
zhangjinchao01 已提交
2125 2126
@config_layer('blockexpand')
class BlockExpandLayer(LayerBase):
2127
    def __init__(self, name, inputs, **xargs):
Q
qijun 已提交
2128
        super(BlockExpandLayer, self).__init__(
2129
            name, 'blockexpand', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2130 2131
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
Q
qijun 已提交
2132 2133
            parse_block_expand(
                self.inputs[input_index].block_expand, input_layer.name,
Z
zhangjinchao01 已提交
2134
                self.config.inputs[input_index].block_expand_conf)
Q
qijun 已提交
2135 2136 2137 2138 2139 2140
            block_expand_conf = self.config.inputs[
                input_index].block_expand_conf
            self.set_layer_size(block_expand_conf.block_x *
                                block_expand_conf.block_y *
                                block_expand_conf.channels)

Z
zhangjinchao01 已提交
2141

2142 2143
@config_layer('maxout')
class MaxOutLayer(LayerBase):
Q
qijun 已提交
2144 2145 2146
    def __init__(self, name, inputs, **xargs):
        super(MaxOutLayer, self).__init__(
            name, 'maxout', 0, inputs=inputs, **xargs)
2147 2148
        input_layer = self.get_input_layer(0)
        maxout_conf = self.config.inputs[0].maxout_conf
L
Luo Tao 已提交
2149
        parse_maxout(self.inputs[0].maxout, input_layer.name, maxout_conf)
L
Luo Tao 已提交
2150 2151 2152
        out_channels = maxout_conf.image_conf.channels / maxout_conf.groups
        self.set_cnn_layer(name, g_layer_map[input_layer.name].height,
                           g_layer_map[input_layer.name].width, out_channels)
Q
qijun 已提交
2153

2154

D
dangqingqing 已提交
2155 2156 2157 2158
@config_layer('row_conv')
class RowConvLayer(LayerBase):
    def __init__(self, name, inputs, context_length, **xargs):
        super(RowConvLayer, self).__init__(
2159
            name, 'row_conv', 0, inputs=inputs, **xargs)
D
dangqingqing 已提交
2160 2161
        config_assert(
            len(self.inputs) == 1,
2162
            'row convolution layer must have one and only one input.')
D
dangqingqing 已提交
2163 2164 2165 2166 2167 2168 2169 2170 2171
        input_layer = self.get_input_layer(0)
        row_conv_conf = self.config.inputs[0].row_conv_conf
        row_conv_conf.context_length = context_length
        self.set_layer_size(input_layer.size)
        psize = context_length * input_layer.size
        dims = [context_length, input_layer.size]
        self.create_input_parameter(0, psize, dims)


Z
zhangjinchao01 已提交
2172 2173 2174 2175
# key: cost type
# value: cost class
g_cost_map = {}

Q
qijun 已提交
2176

Z
zhangjinchao01 已提交
2177 2178 2179
# define a cost layer without any parameters
def define_cost(class_name, cost_type):
    def init(cls, name, inputs, device=None, coeff=1.):
Q
qijun 已提交
2180 2181
        super(type(cls), cls).__init__(
            name, cost_type, 1, inputs, device=device, coeff=coeff)
Z
zhangjinchao01 已提交
2182

Q
qijun 已提交
2183
    cls = type(class_name, (LayerBase, ), dict(__init__=init))
Z
zhangjinchao01 已提交
2184 2185 2186
    global g_cost_map
    g_cost_map[cost_type] = cls

Q
qijun 已提交
2187

Z
zhangjinchao01 已提交
2188 2189 2190 2191 2192 2193 2194 2195
define_cost('MultiClassCrossEntropy', 'multi-class-cross-entropy')
define_cost('RankingCost', 'rank-cost')
define_cost('AucValidation', 'auc-validation')
define_cost('PnpairValidation', 'pnpair-validation')
define_cost('SumOfSquaresCostLayer', 'square_error')
define_cost('MultiBinaryLabelCrossEntropy', 'multi_binary_label_cross_entropy')
define_cost('SoftBinaryClassCrossEntropy', 'soft_binary_class_cross_entropy')
define_cost('HuberTwoClass', 'huber')
X
xuwei06 已提交
2196
define_cost('SumCost', 'sum_cost')
D
dangqingqing 已提交
2197
define_cost('SmoothL1Cost', 'smooth_l1')
Z
zhangjinchao01 已提交
2198

Q
qijun 已提交
2199

Z
zhangjinchao01 已提交
2200 2201
@config_layer('hsigmoid')
class HierarchicalSigmoidLayer(LayerBase):
Q
qijun 已提交
2202
    def __init__(self, name, num_classes, inputs, device=None, bias=True):
Z
zhangjinchao01 已提交
2203 2204
        super(HierarchicalSigmoidLayer, self).__init__(
            name, 'hsigmoid', 1, inputs=inputs, device=device)
Q
qijun 已提交
2205 2206 2207
        config_assert(
            len(self.inputs) >= 2,
            'HierarchicalSigmoidLayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2208 2209 2210 2211 2212 2213 2214 2215
        self.config.num_classes = num_classes
        for input_index in xrange(len(self.inputs) - 1):
            input_layer = self.get_input_layer(input_index)
            psize = (num_classes - 1) * input_layer.size
            dims = [num_classes - 1, input_layer.size]
            self.create_input_parameter(input_index, psize, dims)
        self.create_bias_parameter(bias, num_classes - 1)

Q
qijun 已提交
2216

Z
zhangjinchao01 已提交
2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240
'''
lambdaCost for lambdaRank LTR approach

Usage:
  Example: Layer(name = "cost", type = "lambda_cost", NDCG_num = 8,
             max_sort_size = -1, inputs = ["output", "score"])

  Input data: Samples of the same query should be loaded as a sequence,
          by ProtoDataProvider or PyDataProvider etc.. User should provide
          scores for each sample. The score slot should be the 2nd
          input of lambdaRank layer.

  NDCG_num = the size of NDCG, e.g., 5 for NDCG@5.
    Note: NDCG_num must be less than or equal to the minimum
          size of lists.

  max_sort_size = the size of partial sorting in calculating gradient.
    Note: If max_sort_size = -1, then for each list, the algorithm will
          sort the entire list to get gradient.
          In other cases, max_sort_size must be greater than or equal
          to NDCG_num.
          max_sort_size can be greater than the size of a list, in which
          case the algorithm will sort the entire list to get gradient.
'''
Q
qijun 已提交
2241 2242


Z
zhangjinchao01 已提交
2243 2244
@config_layer('lambda_cost')
class LambdaCost(LayerBase):
Q
qijun 已提交
2245
    def __init__(self, name, inputs, NDCG_num=5, max_sort_size=-1, device=None):
Z
zhangjinchao01 已提交
2246 2247
        super(LambdaCost, self).__init__(
            name, 'lambda_cost', 1, inputs=inputs, device=device)
Q
qijun 已提交
2248
        config_assert(len(self.inputs) == 2, 'lambdaCost must have 2 inputs')
Z
zhangjinchao01 已提交
2249 2250
        self.config.NDCG_num = NDCG_num
        if max_sort_size != -1:
Q
qijun 已提交
2251 2252 2253
            config_assert(
                NDCG_num <= max_sort_size,
                'NDCG_num must be less than or equal to max_sort_size')
Z
zhangjinchao01 已提交
2254 2255
        self.config.max_sort_size = max_sort_size

Q
qijun 已提交
2256

Z
zhangjinchao01 已提交
2257 2258
@config_layer('nce')
class NCELayer(LayerBase):
Q
qijun 已提交
2259 2260 2261 2262 2263 2264 2265 2266
    def __init__(self,
                 name,
                 num_classes,
                 inputs,
                 num_neg_samples=10,
                 neg_sampling_dist=None,
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2267
        super(NCELayer, self).__init__(name, 'nce', 1, inputs=inputs, **xargs)
Q
qijun 已提交
2268 2269
        config_assert(
            len(self.inputs) >= 2, 'NCELayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2270 2271
        self.config.num_classes = num_classes
        if neg_sampling_dist is not None:
Q
qijun 已提交
2272 2273 2274 2275
            config_assert(
                len(neg_sampling_dist) == num_classes,
                'len(neg_sampling_dist)(%s) is not same as num_classes (%s)' %
                (len(neg_sampling_dist), num_classes))
Z
zhangjinchao01 已提交
2276
            s = sum(neg_sampling_dist)
Q
qijun 已提交
2277 2278 2279
            config_assert(
                abs(s - 1) < 1e-5,
                'The sum of neg_sampling_dist (%s) is not 1' % s)
Z
zhangjinchao01 已提交
2280 2281 2282 2283 2284

            self.config.neg_sampling_dist.extend(neg_sampling_dist)

        self.config.num_neg_samples = num_neg_samples
        num_real_inputs = len(self.inputs) - 1
Q
qijun 已提交
2285
        input_layer = self.get_input_layer(num_real_inputs)
Z
zhangjinchao01 已提交
2286 2287 2288 2289
        config_assert(input_layer.type == 'data',
                      'Expecting the last input layer of an nce layer to be '
                      'a data layer')

Q
qijun 已提交
2290 2291
        if (num_real_inputs > 1 and input_layer.size == 1 and
                self.get_input_layer(num_real_inputs - 1).type == 'data'):
Z
zhangjinchao01 已提交
2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
            # This input layer is assumed to be a sample weight layer
            num_real_inputs -= 1

        for input_index in xrange(num_real_inputs):
            input_layer = self.get_input_layer(input_index)
            psize = num_classes * input_layer.size
            dims = [num_classes, input_layer.size]
            self.create_input_parameter(input_index, psize, dims)
        self.create_bias_parameter(bias, num_classes)


@config_layer('addto')
class AddToLayer(LayerBase):
Q
qijun 已提交
2305
    def __init__(self, name, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
2306 2307
        super(AddToLayer, self).__init__(
            name, 'addto', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2308
        config_assert(len(inputs) > 0, 'inputs cannot be empty for AddToLayer')
Z
zhangjinchao01 已提交
2309 2310 2311 2312 2313
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
2314

Z
zhangjinchao01 已提交
2315 2316
@config_layer('agent')
class AgentLayer(LayerBase):
Q
qijun 已提交
2317 2318 2319 2320
    def __init__(self, name, size, device=None):
        super(AgentLayer, self).__init__(
            name, 'agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2321 2322 2323

@config_layer('gather_agent')
class GatherAgentLayer(LayerBase):
Q
qijun 已提交
2324
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2325 2326 2327
        super(GatherAgentLayer, self).__init__(
            name, 'gather_agent', size, inputs=[], device=device)

Q
qijun 已提交
2328

Z
zhangjinchao01 已提交
2329 2330
@config_layer('scatter_agent')
class ScatterAgentLayer(LayerBase):
Q
qijun 已提交
2331
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2332 2333 2334
        super(ScatterAgentLayer, self).__init__(
            name, 'scatter_agent', size, inputs=[], device=device)

Q
qijun 已提交
2335

Z
zhangjinchao01 已提交
2336 2337
@config_layer('multiplex')
class MultiplexLayer(LayerBase):
Q
qijun 已提交
2338 2339 2340 2341 2342
    def __init__(self, name, inputs, size, device=None):
        super(MultiplexLayer, self).__init__(
            name, 'multiplex', size, inputs=inputs, device=device)
        config_assert(
            len(inputs) > 2, 'MultiplexLayer should have more than 2 inputs.')
Z
zhangjinchao01 已提交
2343
        for i in range(1, len(inputs)):
Q
qijun 已提交
2344 2345 2346 2347 2348
            config_assert(
                self.get_input_layer(i).size == size,
                "All the input layers except the first one should"
                "have the same size as the MultiplexLayer.")

Z
zhangjinchao01 已提交
2349 2350

@config_func
2351 2352 2353 2354
def Link(name, has_subseq=False):
    """
    Still keeping has_subseq for backward compatibility
    """
Z
zhangjinchao01 已提交
2355 2356 2357 2358
    link_config = LinkConfig()
    link_config.link_name = name
    return link_config

Q
qijun 已提交
2359

Z
zhangjinchao01 已提交
2360 2361
# memory for recurrent layer group.
# *name* and *size* are actual layer's name and size.
2362 2363 2364 2365
# If *name* is None, need to provide *memory_name* and need to use
# SetMemoryInput() later to specify the layer which this memory remembers.
#
# return the name of the memory,
Z
zhangjinchao01 已提交
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376
# use this name if you assign the memory as other layer's input
#
# boot frame of memory is zeroed by default,
# or initialize by boot layer output if *boot_layer* set,
# or initialize by trainable bias if *boot_bias* set,
# or initialize by a constant id if *boot_with_const_id* set
#
# Memory can be a sequence if *is_sequence* set, this type of memory
# can only be initailized by a *boot_layer* which is a sequence.
#
@config_func
2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
def Memory(name,
           size,
           is_sequence=False,
           boot_layer=None,
           boot_bias=False,
           boot_bias_active_type="",
           boot_with_const_id=None,
           memory_name=None):
    if not memory_name:
        config_assert(name is not None, "name needs cannot be None")
        memory_name = name + "+delay1"
    agent_name = memory_name
2389
    agent_layer = AgentLayer(agent_name, size)
Z
zhangjinchao01 已提交
2390
    config_assert(g_current_submodel.is_recurrent_layer_group,
Q
qijun 已提交
2391
                  'Memory should be used in recurrent layer group only')
Z
zhangjinchao01 已提交
2392
    memory = g_current_submodel.memories.add()
2393 2394
    if name is not None:
        memory.layer_name = MakeLayerNameInSubmodel(name)
Z
zhangjinchao01 已提交
2395
    memory.link_name = MakeLayerNameInSubmodel(agent_name)
Q
qijun 已提交
2396
    options = sum((boot_layer is not None, bool(boot_bias),
Z
zhangjinchao01 已提交
2397
                   boot_with_const_id is not None))
Q
qijun 已提交
2398 2399 2400 2401
    config_assert(
        options <= 1,
        'take one option at most from boot_layer, boot_bias, or boot_with_const_id'
    )
Z
zhangjinchao01 已提交
2402 2403 2404
    if boot_layer is not None:
        boot_layer = MakeLayerNameInParentSubmodel(boot_layer)
        config_assert(boot_layer in g_layer_map,
Q
qijun 已提交
2405 2406
                      'boot_layer "%s" does not correspond to a layer name' %
                      boot_layer)
Z
zhangjinchao01 已提交
2407 2408 2409
        memory.boot_layer_name = boot_layer
    elif boot_bias:
        memory.boot_bias_parameter_name = agent_layer.create_bias_parameter(
Q
qijun 已提交
2410
            boot_bias, size, for_self=False)
Z
zhangjinchao01 已提交
2411 2412 2413 2414 2415
        memory.boot_bias_active_type = boot_bias_active_type
    elif boot_with_const_id is not None:
        memory.boot_with_const_id = boot_with_const_id
    return agent_name

Q
qijun 已提交
2416

2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427
@config_func
def SetMemoryInput(memory_name, layer_name):
    memory_name = MakeLayerNameInSubmodel(memory_name)
    layer_name = MakeLayerNameInSubmodel(layer_name)
    for mem in g_current_submodel.memories:
        if mem.link_name == memory_name:
            mem.layer_name = layer_name
            return
    logger.fatal("Nonexistent memory name: " + memory_name)


Z
zhangjinchao01 已提交
2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
# Generator for recurrent layer group, to use it:
#  1. define a id layer as output of layer group
#  2. define a memory of this id layer, and assign a boot id(begin of sequence)
#  3. define a eos check layer and fill its name in generator's *eos_layer_name*
# Sequence generation will stop when eos check return 1 or *max_num_frames* reached.
# If *beam_size* is greater than one, generator will use beam search.
#   in beam search, if *num_results_per_sample* set, one sample sequence can output
#   multiple results each with a probility.
@config_func
def Generator(
        max_num_frames,
Q
qijun 已提交
2439 2440 2441 2442
        eos_layer_name="eos_check",
        num_results_per_sample=1,
        beam_size=1,
        log_prob=None, ):
Z
zhangjinchao01 已提交
2443 2444 2445 2446 2447 2448 2449 2450 2451
    generator_config = GeneratorConfig()
    generator_config.max_num_frames = max_num_frames
    generator_config.eos_layer_name = eos_layer_name
    generator_config.num_results_per_sample = num_results_per_sample
    generator_config.beam_size = beam_size
    if log_prob is not None:
        generator_config.log_prob = log_prob
    return generator_config

Q
qijun 已提交
2452

Z
zhangjinchao01 已提交
2453 2454
@config_layer('expand')
class ExpandLayer(LayerBase):
2455
    def __init__(self, name, inputs, trans_type='non-seq', bias=False, **xargs):
Q
qijun 已提交
2456
        super(ExpandLayer, self).__init__(
2457
            name, 'expand', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2458 2459 2460 2461 2462 2463 2464 2465
        config_assert(
            len(self.inputs) == 2, 'ExpandLayer takes 2 and only 2 inputs')
        self.config.trans_type = trans_type
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
        self.set_layer_size(self.get_input_layer(0).size)
        self.create_bias_parameter(bias, self.config.size)

Z
zhangjinchao01 已提交
2466 2467 2468

@config_layer('featmap_expand')
class FeatMapExpandLayer(LayerBase):
X
xuwei06 已提交
2469 2470 2471 2472 2473
    def __init__(self,
                 name,
                 inputs,
                 num_filters=None,
                 as_row_vector=True,
X
xuwei06 已提交
2474 2475
                 bias=False,
                 **xargs):
Q
qijun 已提交
2476
        super(FeatMapExpandLayer, self).__init__(
X
xuwei06 已提交
2477
            name, 'featmap_expand', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2478 2479 2480
        config_assert(
            len(self.inputs) == 1, 'ExpandLayer takes 1 and only 1 inputs')
        if num_filters is not None:
Z
zhangjinchao01 已提交
2481
            self.config.num_filters = num_filters
Q
qijun 已提交
2482
        else:
Z
zhangjinchao01 已提交
2483
            logger.fatal("FeatMapExpandLayer must specify num_filters.")
X
xuwei06 已提交
2484 2485
        if not as_row_vector:
            self.config.user_arg = "as_col_vec"
Q
qijun 已提交
2486
        self.set_layer_size(self.get_input_layer(0).size * num_filters)
Z
zhangjinchao01 已提交
2487 2488 2489 2490


@config_layer('max')
class MaxLayer(LayerBase):
Q
qijun 已提交
2491 2492 2493 2494 2495
    def __init__(self,
                 name,
                 inputs,
                 trans_type='non-seq',
                 bias=False,
2496
                 output_max_index=None,
2497
                 stride=-1,
2498
                 **xargs):
2499
        super(MaxLayer, self).__init__(name, 'max', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2500
        config_assert(len(self.inputs) == 1, 'MaxLayer must have 1 input')
2501 2502
        if trans_type == 'seq':
            config_assert(stride == -1, 'subseq does not support stride window')
Q
qijun 已提交
2503
        self.config.trans_type = trans_type
2504
        self.config.seq_pool_stride = stride
Z
zhangjinchao01 已提交
2505 2506 2507 2508
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)
        self.create_bias_parameter(bias, self.config.size)
2509 2510
        if output_max_index is not None:
            self.config.output_max_index = output_max_index
Z
zhangjinchao01 已提交
2511 2512 2513 2514


@config_layer('maxid')
class MaxIdLayer(LayerBase):
Q
qijun 已提交
2515
    def __init__(self, name, inputs, beam_size=None, device=None):
Z
zhangjinchao01 已提交
2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532
        super(MaxIdLayer, self).__init__(
            name, 'maxid', 0, inputs=inputs, device=device)
        config_assert(len(self.inputs) == 1, 'MaxIdLayer must have 1 input')
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)

        if beam_size is None:
            global g_current_submodel
            if g_current_submodel.HasField("generator"):
                self.config.beam_size = g_current_submodel.generator.beam_size
        else:
            self.config.beam_size = beam_size


@config_layer('eos_id')
class EosIdLayer(LayerBase):
Q
qijun 已提交
2533
    def __init__(self, name, inputs, eos_id, device=None):
Z
zhangjinchao01 已提交
2534 2535 2536
        super(EosIdLayer, self).__init__(
            name, 'eos_id', 0, inputs=inputs, device=device)
        config_assert(len(self.inputs) == 1, 'EosIdLayer must have 1 input')
Q
qijun 已提交
2537
        self.set_layer_size(2)  # boolean output
Z
zhangjinchao01 已提交
2538 2539
        self.config.eos_id = eos_id

Q
qijun 已提交
2540

Z
zhangjinchao01 已提交
2541 2542
@config_layer('seqlastins')
class SequenceLastInstanceLayer(LayerBase):
Q
qijun 已提交
2543 2544 2545 2546
    def __init__(self,
                 name,
                 inputs,
                 trans_type='non-seq',
2547
                 bias=False,
2548
                 stride=-1,
2549
                 **xargs):
Q
qijun 已提交
2550
        super(SequenceLastInstanceLayer, self).__init__(
X
xuwei06 已提交
2551
            name, 'seqlastins', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2552 2553
        config_assert(
            len(inputs) == 1, 'SequenceLastInstanceLayer must have 1 input')
2554
        if trans_type == 'seq':
L
Luo Tao 已提交
2555
            config_assert(stride == -1, 'subseq does not support stride window')
Q
qijun 已提交
2556
        self.config.trans_type = trans_type
2557 2558
        self.config.seq_pool_stride = stride
        self.set_layer_size(self.get_input_layer(0).size)
Z
zhangjinchao01 已提交
2559 2560
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
2561

Z
zhangjinchao01 已提交
2562 2563
@config_layer('seqfirstins')
class SequenceFirstInstanceLayer(SequenceLastInstanceLayer):
2564 2565 2566 2567 2568
    def __init__(self,
                 name,
                 inputs,
                 trans_type='non-seq',
                 bias=False,
2569
                 stride=-1,
2570
                 **xargs):
Q
qijun 已提交
2571
        super(SequenceFirstInstanceLayer, self).__init__(
2572 2573 2574 2575 2576 2577
            name,
            inputs=inputs,
            trans_type=trans_type,
            bias=bias,
            stride=stride,
            **xargs)
Z
zhangjinchao01 已提交
2578 2579
        self.config.select_first = True

Q
qijun 已提交
2580

Z
zhangjinchao01 已提交
2581 2582
@config_layer('seqconcat')
class SequenceConcatLayer(LayerBase):
X
xuwei06 已提交
2583
    def __init__(self, name, inputs, bias=False, **xargs):
Q
qijun 已提交
2584
        super(SequenceConcatLayer, self).__init__(
X
xuwei06 已提交
2585
            name, 'seqconcat', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2586 2587
        config_assert(
            len(inputs) == 2, 'SequenceConcatLayer must have 2 inputs')
Z
zhangjinchao01 已提交
2588 2589 2590 2591 2592
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
2593

Z
zhangjinchao01 已提交
2594 2595
@config_layer('seqreshape')
class SequenceReshapeLayer(LayerBase):
X
xuwei06 已提交
2596
    def __init__(self, name, size, inputs, bias=False, **xargs):
Q
qijun 已提交
2597
        super(SequenceReshapeLayer, self).__init__(
X
xuwei06 已提交
2598
            name, 'seqreshape', size, inputs=inputs, **xargs)
Q
qijun 已提交
2599 2600
        config_assert(
            len(inputs) == 1, 'SequenceReshapeLayer must have 1 inputs')
Z
zhangjinchao01 已提交
2601 2602 2603
        self.set_layer_size(size)
        self.create_bias_parameter(bias, size)

Q
qijun 已提交
2604

Z
zhangjinchao01 已提交
2605 2606
@config_layer('subseq')
class SubSequenceLayer(LayerBase):
X
xuwei06 已提交
2607
    def __init__(self, name, inputs, bias=False, **xargs):
Q
qijun 已提交
2608
        super(SubSequenceLayer, self).__init__(
X
xuwei06 已提交
2609
            name, 'subseq', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2610 2611 2612 2613 2614 2615
        config_assert(len(inputs) == 3, 'SubSequenceLayer must have 3 inputs')
        input_layer0 = self.get_input_layer(0)
        size = input_layer0.size
        self.set_layer_size(size)
        self.create_bias_parameter(bias, size)

Q
qijun 已提交
2616

Z
zhangjinchao01 已提交
2617 2618
@config_layer('out_prod')
class OuterProdLayer(LayerBase):
Q
qijun 已提交
2619 2620 2621
    def __init__(self, name, inputs, device=None):
        super(OuterProdLayer, self).__init__(
            name, 'out_prod', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2622 2623 2624 2625 2626
        config_assert(len(inputs) == 2, 'OuterProdLayer must have 2 inputs')
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)
        self.set_layer_size(input_layer0.size * input_layer1.size)

Q
qijun 已提交
2627

Z
zhangjinchao01 已提交
2628 2629
@config_layer('power')
class PowerLayer(LayerBase):
Q
qijun 已提交
2630 2631 2632
    def __init__(self, name, inputs, device=None):
        super(PowerLayer, self).__init__(
            name, 'power', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2633 2634 2635 2636
        config_assert(len(inputs) == 2, 'PowerLayer must have 2 inputs')
        input_layer1 = self.get_input_layer(1)
        self.set_layer_size(input_layer1.size)
        input_layer0 = self.get_input_layer(0)
Q
qijun 已提交
2637 2638 2639
        config_assert(1 == input_layer0.size,
                      'The left input is the exponent and should be of size 1')

Z
zhangjinchao01 已提交
2640 2641 2642

@config_layer('slope_intercept')
class SlopeInterceptLayer(LayerBase):
Q
qijun 已提交
2643 2644 2645
    def __init__(self, name, inputs, slope=1.0, intercept=0.0, device=None):
        super(SlopeInterceptLayer, self).__init__(
            name, 'slope_intercept', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2646 2647 2648 2649 2650 2651
        self.config.slope = slope
        self.config.intercept = intercept
        config_assert(len(inputs) == 1, 'SlopeInterceptLayer must have 1 input')
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2652

Z
zhangjinchao01 已提交
2653 2654
@config_layer('scaling')
class ScalingLayer(LayerBase):
Q
qijun 已提交
2655 2656 2657
    def __init__(self, name, inputs, device=None):
        super(ScalingLayer, self).__init__(
            name, 'scaling', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2658 2659 2660 2661
        config_assert(len(inputs) == 2, 'ScalingLayer must have 2 inputs')
        input_layer1 = self.get_input_layer(1)
        self.set_layer_size(input_layer1.size)
        input_layer0 = self.get_input_layer(0)
Q
qijun 已提交
2662 2663 2664
        config_assert(1 == input_layer0.size,
                      'The left input should be of size 1')

Z
zhangjinchao01 已提交
2665 2666 2667

@config_layer('conv_shift')
class ConvShiftLayer(LayerBase):
Q
qijun 已提交
2668 2669 2670
    def __init__(self, name, inputs, device=None):
        super(ConvShiftLayer, self).__init__(
            name, 'conv_shift', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2671 2672 2673 2674
        config_assert(len(inputs) == 2, 'ConvShiftLayer must have 2 inputs')
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2675

Z
zhangjinchao01 已提交
2676 2677
@config_layer('convex_comb')
class ConvexCombinationLayer(LayerBase):
Q
qijun 已提交
2678
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
2679
        super(ConvexCombinationLayer, self).__init__(
Q
qijun 已提交
2680 2681 2682
            name, 'convex_comb', size, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 2, 'ConvexCombinationLayer must have 2 inputs')
2683 2684 2685
        config_assert(
            size * self.get_input_layer(0).size == self.get_input_layer(1).size,
            'Wrong input size for ConvexCombinationLayer')
Z
zhangjinchao01 已提交
2686 2687
        self.set_layer_size(size)

Q
qijun 已提交
2688

Z
zhangjinchao01 已提交
2689 2690
@config_layer('interpolation')
class InterpolationLayer(LayerBase):
Q
qijun 已提交
2691
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2692 2693
        super(InterpolationLayer, self).__init__(
            name, 'interpolation', 0, inputs=inputs, device=device)
Q
qijun 已提交
2694 2695
        config_assert(
            len(self.inputs) == 3, 'InterpolationLayer must have 3 inputs')
Z
zhangjinchao01 已提交
2696 2697 2698 2699 2700 2701 2702 2703
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)
        input_layer2 = self.get_input_layer(2)
        self.set_layer_size(input_layer1.size)
        config_assert(input_layer0.size == 1, 'weight should be of size 1')
        config_assert(input_layer1.size == input_layer2.size,
                      'the two vector inputs should be of the same size')

Q
qijun 已提交
2704

L
liaogang 已提交
2705 2706
@config_layer('bilinear_interp')
class BilinearInterpLayer(LayerBase):
Q
qijun 已提交
2707
    def __init__(self, name, inputs, **xargs):
L
liaogang 已提交
2708
        super(BilinearInterpLayer, self).__init__(
L
liaogang 已提交
2709
            name, 'bilinear_interp', 0, inputs=inputs, **xargs)
L
liaogang 已提交
2710
        input_layer = self.get_input_layer(0)
L
Luo Tao 已提交
2711 2712 2713 2714
        conf = self.config.inputs[0].bilinear_interp_conf
        parse_bilinear(self.inputs[0].bilinear_interp, input_layer.name, conf)
        self.set_cnn_layer(name, conf.out_size_y, conf.out_size_x,
                           conf.image_conf.channels)
Q
qijun 已提交
2715

L
liaogang 已提交
2716

Z
zhangjinchao01 已提交
2717 2718
@config_layer('sum_to_one_norm')
class SumToOneNormLayer(LayerBase):
Q
qijun 已提交
2719
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2720
        super(SumToOneNormLayer, self).__init__(
Q
qijun 已提交
2721 2722 2723
            name, 'sum_to_one_norm', 0, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 1, 'SumToOneNormLayer must have 1 input')
Z
zhangjinchao01 已提交
2724 2725 2726
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2727

Z
zhangjinchao01 已提交
2728 2729
@config_layer('cos_vm')
class CosSimVecMatLayer(LayerBase):
Q
qijun 已提交
2730
    def __init__(self, name, size, inputs, cos_scale=1.0, device=None):
Z
zhangjinchao01 已提交
2731
        super(CosSimVecMatLayer, self).__init__(
Q
qijun 已提交
2732
            name, 'cos_vm', size, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2733
        self.config.cos_scale = cos_scale
Q
qijun 已提交
2734 2735
        config_assert(
            len(self.inputs) == 2, 'CosSimVecMatLayer must have 2 inputs')
2736 2737 2738
        config_assert(
            size * self.get_input_layer(0).size == self.get_input_layer(1).size,
            'Wrong input size for CosSimVecMatLayer')
Z
zhangjinchao01 已提交
2739

Q
qijun 已提交
2740

Z
zhangjinchao01 已提交
2741 2742
@config_layer('sampling_id')
class SamplingIdLayer(LayerBase):
Q
qijun 已提交
2743
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2744 2745
        super(SamplingIdLayer, self).__init__(
            name, 'sampling_id', 0, inputs=inputs, device=device)
Q
qijun 已提交
2746 2747
        config_assert(
            len(self.inputs) == 1, 'SamplingIdLayer must have 1 input')
Z
zhangjinchao01 已提交
2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)


# AverageLayer: "average" for each sample within a sequence.
# average_stratrgy: set to one of the following:
# 'average': plain average.
# 'sum': sum each sample instead of average (which is divide by sample_num).
# 'squarerootn': sum each sample, but divide by sqrt(sample_num).
@config_layer('average')
class AverageLayer(LayerBase):
Q
qijun 已提交
2760 2761 2762 2763 2764
    def __init__(self,
                 name,
                 inputs,
                 average_strategy='average',
                 trans_type='non-seq',
2765
                 bias=False,
2766
                 stride=-1,
2767
                 **xargs):
Q
qijun 已提交
2768
        super(AverageLayer, self).__init__(
X
xuwei06 已提交
2769
            name, 'average', 0, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2770
        self.config.average_strategy = average_strategy
2771 2772
        if trans_type == 'seq':
            config_assert(stride == -1, 'subseq does not support stride window')
Q
qijun 已提交
2773
        self.config.trans_type = trans_type
2774
        self.config.seq_pool_stride = stride
Z
zhangjinchao01 已提交
2775 2776 2777 2778 2779 2780
        config_assert(len(inputs) == 1, 'AverageLayer must have 1 input')
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            self.set_layer_size(input_layer.size)
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
2781

Z
zhangjinchao01 已提交
2782 2783
@config_layer('cos')
class CosSimLayer(LayerBase):
2784
    def __init__(self, name, inputs, cos_scale=1, device=None):
Z
zhangjinchao01 已提交
2785 2786 2787 2788 2789 2790
        super(CosSimLayer, self).__init__(
            name, 'cos', 1, inputs=inputs, device=device)
        config_assert(len(self.inputs) == 2, 'CosSimLayer must have 2 inputs')
        config_assert(
            self.get_input_layer(0).size == self.get_input_layer(1).size,
            'inputs of CosSimLayer must have same dim')
2791
        self.config.cos_scale = cos_scale
Z
zhangjinchao01 已提交
2792 2793 2794 2795


@config_layer('tensor')
class TensorLayer(LayerBase):
2796
    def __init__(self, name, size, inputs, bias=True, **xargs):
Q
qijun 已提交
2797
        super(TensorLayer, self).__init__(
2798
            name, 'tensor', size, inputs=inputs, **xargs)
Z
zhangjinchao01 已提交
2799 2800
        config_assert(len(self.inputs) == 2, 'TensorLayer must have 2 inputs')
        config_assert(size > 0, 'size must be positive')
Q
qijun 已提交
2801 2802
        config_assert(inputs[1].parameter_name == None,
                      'second parameter should be None.')
Z
zhangjinchao01 已提交
2803 2804 2805 2806 2807 2808 2809 2810 2811 2812
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)
        psize = size * input_layer0.size * input_layer1.size
        dims = [input_layer0.size, input_layer1.size, size]
        self.create_input_parameter(0, psize, dims)
        self.create_bias_parameter(bias, size)


@config_layer('mixed')
class MixedLayer(LayerBase):
C
caoying03 已提交
2813
    def __init__(self, name, inputs, size=0, bias=True, **xargs):
Z
zhangjinchao01 已提交
2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830
        config_assert(inputs, 'inputs cannot be empty')
        super(MixedLayer, self).__init__(
            name, 'mixed', size, inputs=inputs, **xargs)
        operator_input_index = []
        for operator in self.operators:
            operator_conf = operator.operator_conf
            for i in xrange(1, len(operator.input_layer_names)):
                input_index = len(self.config.inputs)
                operator_conf.input_indices.append(input_index)
                input_config = Input(operator.input_layer_names[i])
                self.inputs.append(input_config)
                layer_input = self.config.inputs.add()
                layer_input.input_layer_name = input_config.input_layer_name
            for input_index in operator_conf.input_indices:
                input_layer = self.get_input_layer(input_index)
                operator_conf.input_sizes.append(input_layer.size)
                operator_input_index.append(input_index)
2831
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2832 2833 2834
                size = operator.calc_output_size(operator_conf.input_sizes)
                if size != 0:
                    self.set_layer_size(size)
2835
            else:
2836 2837
                sz = operator.calc_output_size(operator_conf.input_sizes)
                if sz != 0:
Q
qijun 已提交
2838 2839 2840 2841
                    config_assert(
                        sz == self.config.size,
                        "different inputs have different size: %s vs. %s" %
                        (sz, self.config.size))
Z
zhangjinchao01 已提交
2842 2843 2844 2845
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            input = self.inputs[input_index]
            if input_index not in operator_input_index:
Q
qijun 已提交
2846 2847 2848
                config_assert(
                    isinstance(input, Projection),
                    "input should be projection or operation")
2849
            if self.config.size == 0 and isinstance(input, Projection):
Z
zhangjinchao01 已提交
2850 2851 2852
                size = input.calc_output_size(input_layer)
                if size != 0:
                    self.set_layer_size(size)
2853
            elif isinstance(input, Projection):
Q
qijun 已提交
2854 2855 2856 2857 2858 2859
                sz = input.calc_output_size(input_layer)
                if sz != 0:
                    config_assert(
                        sz == self.config.size,
                        "different inputs have different size: %s vs. %s" %
                        (sz, self.config.size))
Z
zhangjinchao01 已提交
2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870
        config_assert(size != 0, "size is not set")

        for input_index in xrange(len(self.inputs)):
            input = self.inputs[input_index]
            if isinstance(input, Projection):
                input_layer = self.get_input_layer(input_index)
                input.proj_conf.input_size = input_layer.size
                input.proj_conf.output_size = size

                input_config = self.config.inputs[input_index]
                input_config.proj_conf.CopyFrom(input.proj_conf)
Q
qijun 已提交
2871 2872
                input_config.proj_conf.name = gen_parameter_name(name,
                                                                 input_index)
Z
zhangjinchao01 已提交
2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
                psize = input.calc_parameter_size(input_layer.size, size)
                dims = input.calc_parameter_dims(input_layer.size, size)
                self.create_input_parameter(input_index, psize, dims)

        for operator in self.operators:
            operator_conf = operator.operator_conf
            operator_conf.output_size = self.config.size
            operator.check_dims()
            record_operator_conf = self.config.operator_confs.add()
            record_operator_conf.CopyFrom(operator_conf)

2884 2885 2886 2887 2888 2889
        psize = self.config.size
        if isinstance(self.inputs[0], ConvProjection):
            self.config.shared_biases = True
            psize = 0
            for input in self.inputs:
                psize += input.calc_bias_size()
Z
zhangjinchao01 已提交
2890

2891 2892 2893
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
Z
zhangjinchao01 已提交
2894

Q
qijun 已提交
2895

Z
zhangjinchao01 已提交
2896 2897
# like MixedLayer, but no bias parameter
@config_func
Q
qijun 已提交
2898
def ExpressionLayer(name, inputs, **xargs):
Z
zhangjinchao01 已提交
2899 2900
    MixedLayer(name, inputs, bias=False, **xargs)

Q
qijun 已提交
2901

Z
zhangjinchao01 已提交
2902 2903
@config_layer('concat')
class ConcatenateLayer(LayerBase):
Q
qijun 已提交
2904
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2905
        config_assert(inputs, 'inputs cannot be empty')
2906
        config_assert(not bias, 'ConcatenateLayer cannot support bias.')
Z
zhangjinchao01 已提交
2907 2908 2909 2910 2911 2912
        super(ConcatenateLayer, self).__init__(
            name, 'concat', 0, inputs=inputs, **xargs)
        size = 0
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            input = self.inputs[input_index]
Q
qijun 已提交
2913
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2914 2915 2916 2917
                size += input_layer.size

        self.set_layer_size(size)

Q
qijun 已提交
2918

Z
zhangjinchao01 已提交
2919 2920 2921
# like concat layer, but each input layer was processed by a Projection.
@config_layer('concat2')
class ConcatenateLayer2(LayerBase):
Q
qijun 已提交
2922
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2923 2924 2925
        config_assert(inputs, 'inputs cannot be empty')
        super(ConcatenateLayer2, self).__init__(
            name, 'concat2', 0, inputs=inputs, **xargs)
2926 2927

        if isinstance(self.inputs[0], ConvProjection):
Q
qijun 已提交
2928 2929 2930 2931 2932 2933
            for input_index in xrange(len(self.inputs) - 1):
                input = self.inputs[input_index + 1]
                config_assert(
                    isinstance(input, ConvProjection),
                    "The first input of ConcatenateLayer2 is ConvProjection, "
                    "the other inputs should also be ConvProjection.")
2934

Z
zhangjinchao01 已提交
2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954
        size = 0
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            input = self.inputs[input_index]
            output_size = input.calc_output_size(input_layer)
            config_assert(output_size != 0, "proj output size is not set")
            size += output_size

        self.set_layer_size(size)

        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
            input = self.inputs[input_index]
            input.proj_conf.input_size = input_layer.size
            input.proj_conf.output_size = input.calc_output_size(input_layer)

            input_config = self.config.inputs[input_index]
            input_config.proj_conf.CopyFrom(input.proj_conf)
            input_config.proj_conf.name = gen_parameter_name(name, input_index)
            psize = input.calc_parameter_size(input.proj_conf.input_size,
Q
qijun 已提交
2955
                                              input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2956
            dims = input.calc_parameter_dims(input.proj_conf.input_size,
Q
qijun 已提交
2957
                                             input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2958 2959
            self.create_input_parameter(input_index, psize, dims)

2960 2961 2962 2963 2964 2965 2966
        psize = self.config.size
        if isinstance(self.inputs[0], ConvProjection):
            self.config.shared_biases = True
            psize = 0
            for input in self.inputs:
                psize += input.calc_bias_size()

2967 2968 2969
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
2970

Q
qijun 已提交
2971

Z
zhangjinchao01 已提交
2972 2973
@config_layer('recurrent')
class RecurrentLayer(LayerBase):
Q
qijun 已提交
2974
    def __init__(self, name, inputs, reversed=False, bias=True, **xargs):
Y
Yu Yang 已提交
2975 2976
        super(RecurrentLayer, self).__init__(name, 'recurrent', 0, inputs,
                                             **xargs)
Z
zhangjinchao01 已提交
2977 2978 2979 2980 2981 2982 2983 2984 2985
        config_assert(len(self.inputs) == 1, 'RecurrentLayer must have 1 input')
        input_layer = self.get_input_layer(0)
        size = input_layer.size
        self.set_layer_size(size)
        self.config.reversed = reversed
        dims = [size, size]
        self.create_input_parameter(0, size * size, dims)
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
2986

Z
zhangjinchao01 已提交
2987 2988
@config_layer('lstmemory')
class LstmLayer(LayerBase):
Q
qijun 已提交
2989 2990 2991 2992 2993 2994 2995 2996
    def __init__(self,
                 name,
                 inputs,
                 reversed=False,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2997 2998 2999 3000 3001 3002 3003 3004
        super(LstmLayer, self).__init__(name, 'lstmemory', 0, inputs, **xargs)
        config_assert(len(self.inputs) == 1, 'LstmLayer must have 1 input')
        input_layer = self.get_input_layer(0)
        #check input_layer.size is divided by 4
        config_assert(input_layer.size % 4 == 0, "size % 4 should be 0!")
        size = input_layer.size / 4
        self.set_layer_size(size)
        self.config.reversed = reversed
Q
qijun 已提交
3005
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
3006 3007 3008 3009 3010
        self.config.active_state_type = active_state_type
        self.create_input_parameter(0, size * size * 4, [size, size, 4])
        #bias includes 3 kinds of peephole, 4 + 3 = 7
        self.create_bias_parameter(bias, size * 7)

Q
qijun 已提交
3011

Z
zhangjinchao01 已提交
3012 3013
@config_layer('lstm_step')
class LstmStepLayer(LayerBase):
Q
qijun 已提交
3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
        super(LstmStepLayer, self).__init__(name, 'lstm_step', size, inputs,
                                            **xargs)
Z
zhangjinchao01 已提交
3024 3025 3026
        config_assert(len(inputs) == 2, 'LstmStepLayer must have 2 inputs')
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)
Q
qijun 已提交
3027 3028 3029 3030 3031
        config_assert(input_layer0.size == 4 * size,
                      'input_layer0.size != 4 * layer.size')
        config_assert(input_layer1.size == size,
                      'input_layer1.size != layer.size')
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
3032 3033 3034
        self.config.active_state_type = active_state_type
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
3035

Z
zhangjinchao01 已提交
3036 3037 3038
# get the specific output from the input layer.
@config_layer('get_output')
class GetOutputLayer(LayerBase):
Q
qijun 已提交
3039 3040 3041 3042
    def __init__(self, name, size, inputs):
        super(GetOutputLayer, self).__init__(name, 'get_output', size, inputs)
        config_assert(
            len(self.inputs) == 1, 'GetOutputLayer must have 1 inputs')
Z
zhangjinchao01 已提交
3043 3044 3045 3046
        inputs = self.inputs[0]
        config_assert(inputs.input_layer_argument,
                      'input_layer_argument cannot be empty')

Q
qijun 已提交
3047

Z
zhangjinchao01 已提交
3048 3049
@config_layer('mdlstmemory')
class MDLstmLayer(LayerBase):
Q
qijun 已提交
3050 3051 3052 3053 3054 3055 3056 3057
    def __init__(self,
                 name,
                 inputs,
                 directions=True,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
3058 3059
        super(MDLstmLayer, self).__init__(name, 'mdlstmemory', 0, inputs,
                                          **xargs)
Z
zhangjinchao01 已提交
3060 3061 3062 3063
        config_assert(len(self.inputs) == 1, 'MDLstmLayer must have 1 input')
        input_layer = self.get_input_layer(0)
        dim_num = len(directions)
        #check input_layer.size is divided by (3+dim_num)
Y
Yu Yang 已提交
3064 3065
        config_assert(input_layer.size % (3 + dim_num) == 0,
                      "size % (dim_num) should be 0!")
Q
qijun 已提交
3066
        size = input_layer.size / (3 + dim_num)
Z
zhangjinchao01 已提交
3067
        self.set_layer_size(size)
Q
qijun 已提交
3068
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
3069 3070 3071
        self.config.active_state_type = active_state_type
        for i in xrange(len(directions)):
            self.config.directions.append(int(directions[i]))
Y
Yu Yang 已提交
3072 3073
        self.create_input_parameter(0, size * size * (3 + dim_num),
                                    [size, size, 3 + dim_num])
Z
zhangjinchao01 已提交
3074
        #bias includes 3 kinds of peephole, 3+dim_num+2+dim_num
Q
qijun 已提交
3075 3076
        self.create_bias_parameter(bias, size * (5 + 2 * dim_num))

Z
zhangjinchao01 已提交
3077 3078 3079

@config_layer('gated_recurrent')
class GatedRecurrentLayer(LayerBase):
Q
qijun 已提交
3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090
    def __init__(self,
                 name,
                 inputs,
                 reversed=False,
                 active_gate_type="sigmoid",
                 bias=True,
                 **xargs):
        super(GatedRecurrentLayer, self).__init__(name, 'gated_recurrent', 0,
                                                  inputs, **xargs)
        config_assert(
            len(self.inputs) == 1, 'GatedRecurrentLayer must have 1 input')
Z
zhangjinchao01 已提交
3091 3092 3093 3094 3095 3096
        input_layer = self.get_input_layer(0)
        #check input_layer.size is divided by 3
        config_assert(input_layer.size % 3 == 0, "size % 3 should be 0!")
        size = input_layer.size / 3
        self.set_layer_size(size)
        self.config.reversed = reversed
Q
qijun 已提交
3097
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
3098 3099 3100
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
3101

Z
zhangjinchao01 已提交
3102 3103
@config_layer('gru_step')
class GruStepLayer(LayerBase):
Q
qijun 已提交
3104 3105 3106 3107 3108 3109 3110
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_gate_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
3111 3112
        super(GruStepLayer, self).__init__(name, 'gru_step', size, inputs,
                                           **xargs)
Z
zhangjinchao01 已提交
3113 3114 3115
        config_assert(len(self.inputs) == 2, 'GruStepLayer must have 2 input')
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)
Q
qijun 已提交
3116 3117 3118 3119 3120
        config_assert(input_layer0.size == 3 * size,
                      'input_layer0.size != 3 * layer.size')
        config_assert(input_layer1.size == size,
                      'input_layer1.size != layer.size')
        self.config.active_gate_type = active_gate_type
H
Haonan 已提交
3121
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
Z
zhangjinchao01 已提交
3122 3123
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
3124

Z
zhangjinchao01 已提交
3125 3126 3127 3128 3129 3130 3131
'''
 A layer for calculating the cost of sequential conditional random field model.
 Example: CRFLayer(name="crf_cost", size=label_num,
                   inputs=["output", "label", "weight"])
          where "weight" is optional, one weight for each sequence
 @param coeff: weight of the layer
'''
Q
qijun 已提交
3132 3133


Z
zhangjinchao01 已提交
3134 3135
@config_layer('crf')
class CRFLayer(LayerBase):
Q
qijun 已提交
3136
    def __init__(self, name, size, inputs, coeff=1.0, device=None):
Z
zhangjinchao01 已提交
3137
        super(CRFLayer, self).__init__(name, 'crf', size, inputs, device=device)
Q
qijun 已提交
3138 3139
        config_assert(2 <= len(self.inputs) <= 3,
                      'CRFLayer must have 2 or 3 inputs')
3140
        self.create_input_parameter(0, size * (size + 2), [size + 2, size])
Z
zhangjinchao01 已提交
3141 3142
        self.config.coeff = coeff

Q
qijun 已提交
3143

Z
zhangjinchao01 已提交
3144 3145 3146 3147 3148 3149 3150 3151
'''
 A layer for calculating the decoding sequence of sequential conditional
 random field model.
 The decoding sequence is stored in output_.ids
 If a second input is provided, it is treated as the ground-truth label, and
 this layer will also calculate error, output_.value[i] is 1 for incorrect
 decoding or 0 for correct decoding
'''
Q
qijun 已提交
3152 3153


Z
zhangjinchao01 已提交
3154 3155
@config_layer('crf_decoding')
class CRFDecodingLayer(LayerBase):
Q
qijun 已提交
3156
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
3157 3158 3159 3160 3161
        super(CRFDecodingLayer, self).__init__(
            name, 'crf_decoding', size, inputs, device=device)
        config_assert(
            len(self.inputs) <= 2,
            'CRFDecodingLayer cannot have more than 2 inputs')
3162
        self.create_input_parameter(0, size * (size + 2), [size + 2, size])
Z
zhangjinchao01 已提交
3163

Q
qijun 已提交
3164

Z
zhangjinchao01 已提交
3165 3166
@config_layer('ctc')
class CTCLayer(LayerBase):
Q
qijun 已提交
3167
    def __init__(self, name, size, inputs, norm_by_times=False, device=None):
Z
zhangjinchao01 已提交
3168 3169 3170 3171
        super(CTCLayer, self).__init__(name, 'ctc', size, inputs, device=device)
        self.config.norm_by_times = norm_by_times
        config_assert(len(self.inputs) == 2, 'CTCLayer must have 2 inputs')

Q
qijun 已提交
3172

3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193
@config_layer('warp_ctc')
class WarpCTCLayer(LayerBase):
    def __init__(self,
                 name,
                 size,
                 inputs,
                 blank=0,
                 norm_by_times=False,
                 device=None):
        super(WarpCTCLayer, self).__init__(
            name, 'warp_ctc', size=size, inputs=inputs, device=device)
        self.config.blank = blank
        self.config.norm_by_times = norm_by_times
        config_assert(len(self.inputs) == 2, 'WarpCTCLayer must have 2 inputs')
        input_layer = self.get_input_layer(0)
        config_assert(
            (input_layer.active_type == '' or
             input_layer.active_type == 'linear'),
            "Expecting the active_type of input layer to be linear or null")


Z
zhangjinchao01 已提交
3194 3195
@config_layer('recurrent_layer_group')
class RecurrentLayerGroup(LayerBase):
Q
qijun 已提交
3196
    def __init__(self, name, device=None):
Z
zhangjinchao01 已提交
3197 3198 3199 3200 3201 3202
        super(RecurrentLayerGroup, self).__init__(
            name, 'recurrent_layer_group', 0, inputs=[], device=device)


# Deprecated, use a new layer specific class instead
@config_func
Q
qijun 已提交
3203
def Layer(name, type, **xargs):
Z
zhangjinchao01 已提交
3204 3205 3206 3207
    layers = {}
    layers.update(g_cost_map)
    layers.update(g_layer_type_map)
    layer_func = layers.get(type)
Q
qijun 已提交
3208
    config_assert(layer_func, "layer type '%s' not supported." % type)
X
xuwei06 已提交
3209
    return layer_func(name, **xargs)
Z
zhangjinchao01 已提交
3210

Q
qijun 已提交
3211

Z
zhangjinchao01 已提交
3212
@config_func
Q
qijun 已提交
3213
def ParameterHook(type, **kwargs):
3214
    if type == 'pruning':
Z
zhangjinchao01 已提交
3215 3216
        hook = ParameterUpdaterHookConfig()
        hook.type = type
X
xzl 已提交
3217
        sparsity_ratio = kwargs.get('sparsity_ratio', None)
X
xzl 已提交
3218 3219
        if sparsity_ratio is not None:
            hook.sparsity_ratio = sparsity_ratio
Z
zhangjinchao01 已提交
3220
        return hook
3221 3222 3223 3224
    elif type == 'dpruning':
        hook = ParameterUpdaterHookConfig()
        hook.type = type
        return hook
Z
zhangjinchao01 已提交
3225 3226 3227 3228 3229
    else:
        return None


@config_func
Q
qijun 已提交
3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250
def Parameter(name,
              size,
              device,
              dims,
              learning_rate=None,
              momentum=None,
              decay_rate=None,
              decay_rate_l1=None,
              initial_mean=None,
              initial_std=None,
              initial_strategy=None,
              initial_smart=None,
              num_batches_regularization=None,
              sparse_remote_update=None,
              sparse_update=None,
              gradient_clipping_threshold=None,
              sparse=None,
              format=None,
              need_compact=None,
              is_static=None,
              is_shared=None,
X
xuwei06 已提交
3251 3252
              update_hooks=None,
              initializer=None):
Z
zhangjinchao01 已提交
3253 3254 3255 3256 3257 3258 3259

    config_assert(name not in g_parameter_map,
                  'Duplicated parameter name: ' + name)

    para = g_config.model_config.parameters.add()
    para.name = name
    para.size = size
3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
    if device is not None:
        para.device = int(device)
    para.dims.extend(dims)

    if learning_rate is not None:
        para.learning_rate = float(learning_rate)

    momentum = default(momentum, g_default_momentum)
    if momentum is not None:
        para.momentum = float(momentum)

Z
zhangjinchao01 已提交
3271 3272
    config_assert(not momentum or not decay_rate_l1,
                  "momentum and decay_rate_l1 cannot both be non-zero")
3273 3274 3275 3276 3277

    decay_rate = default(decay_rate, g_default_decay_rate)
    if decay_rate is not None:
        para.decay_rate = decay_rate

Z
zhangjinchao01 已提交
3278 3279 3280 3281
    if decay_rate_l1 is not None:
        para.decay_rate_l1 = decay_rate_l1
    para.initial_std = default(initial_std, g_default_initial_std)
    para.initial_mean = default(initial_mean, g_default_initial_mean)
3282

Q
qijun 已提交
3283 3284
    num_batches_regularization = default(num_batches_regularization,
                                         g_default_num_batches_regularization)
3285 3286 3287
    if num_batches_regularization is not None:
        para.num_batches_regularization = int(num_batches_regularization)

Z
zhangjinchao01 已提交
3288 3289 3290 3291 3292 3293
    if sparse_remote_update is not None:
        para.sparse_remote_update = sparse_remote_update
        if sparse_remote_update:
            g_config.opt_config.use_sparse_remote_updater = True
    if sparse_update is not None:
        para.sparse_update = sparse_update
Q
qijun 已提交
3294 3295
    gradient_clipping_threshold = default(gradient_clipping_threshold,
                                          g_default_gradient_clipping_threshold)
3296 3297
    if gradient_clipping_threshold is not None:
        para.gradient_clipping_threshold = gradient_clipping_threshold
Q
qijun 已提交
3298 3299
    para.initial_strategy = default(initial_strategy,
                                    g_default_initial_strategy)
Z
zhangjinchao01 已提交
3300 3301 3302 3303 3304 3305
    para.initial_smart = default(initial_smart, g_default_initial_smart)
    if para.initial_smart:
        para.initial_mean = 0.
        if len(para.dims) != 0:
            para.initial_std = 1. / math.sqrt(para.dims[0])
        else:
Q
qijun 已提交
3306 3307 3308
            print(
                "Use initial_smart, but dims not set. Initial_smart may not be used in this layer"
            )
Z
zhangjinchao01 已提交
3309 3310 3311 3312
            traceback.print_exc()
            para.initial_std = 1. / math.sqrt(para.size)
    if g_default_compact_func is not None:
        sparse, format, need_compact = g_default_compact_func(para.name)
3313 3314 3315 3316 3317 3318 3319

    if sparse is not None:
        para.is_sparse = sparse
    if format is not None:
        para.format = format
    if need_compact is not None:
        para.need_compact = need_compact
Z
zhangjinchao01 已提交
3320 3321 3322 3323
    if is_static is not None:
        para.is_static = is_static
    config_assert(not para.sparse_remote_update or not para.is_static,
                  "sparse_remote_update and is_static cannot both be true")
3324 3325
    if is_shared is not None:
        para.is_shared = is_shared
Z
zhangjinchao01 已提交
3326 3327 3328 3329 3330

    update_hooks = default(update_hooks, g_default_update_hooks)

    if update_hooks is not None:
        if hasattr(update_hooks, '__call__'):
X
xzl 已提交
3331
            update_hooks = update_hooks()
Z
zhangjinchao01 已提交
3332 3333 3334 3335 3336

        if isinstance(update_hooks, list):
            for hook in update_hooks:
                para.update_hooks.extend([hook])
        else:
X
xzl 已提交
3337
            para.update_hooks.extend([update_hooks])
Z
zhangjinchao01 已提交
3338 3339

    g_parameter_map[name] = para
X
xuwei06 已提交
3340 3341 3342 3343 3344
    if initializer is not None:
        config_assert(
            callable(initializer),
            "parameter initializer should be a callable object")
        g_parameter_initializer_map[name] = initializer
Z
zhangjinchao01 已提交
3345 3346 3347 3348 3349 3350 3351


@config_func
def default_initial_std(val):
    global g_default_initial_std
    g_default_initial_std = val

Q
qijun 已提交
3352

Z
zhangjinchao01 已提交
3353 3354 3355 3356 3357
@config_func
def default_initial_mean(val):
    global g_default_initial_mean
    g_default_initial_mean = val

Q
qijun 已提交
3358

Z
zhangjinchao01 已提交
3359 3360 3361 3362 3363
@config_func
def default_initial_strategy(val):
    global g_default_initial_strategy
    g_default_initial_strategy = val

Q
qijun 已提交
3364

Z
zhangjinchao01 已提交
3365 3366 3367 3368 3369
@config_func
def default_initial_smart(val):
    global g_default_initial_smart
    g_default_initial_smart = val

Q
qijun 已提交
3370

Z
zhangjinchao01 已提交
3371 3372 3373 3374 3375
@config_func
def default_momentum(val):
    global g_default_momentum
    g_default_momentum = val

Q
qijun 已提交
3376

Z
zhangjinchao01 已提交
3377 3378 3379 3380 3381
@config_func
def default_decay_rate(val):
    global g_default_decay_rate
    g_default_decay_rate = val

Q
qijun 已提交
3382

Z
zhangjinchao01 已提交
3383 3384 3385 3386 3387
@config_func
def default_num_batches_regularization(val):
    global g_default_num_batches_regularization
    g_default_num_batches_regularization = val

Q
qijun 已提交
3388

Z
zhangjinchao01 已提交
3389 3390 3391 3392 3393
@config_func
def default_gradient_clipping_threshold(val):
    global g_default_gradient_clipping_threshold
    g_default_gradient_clipping_threshold = val

Q
qijun 已提交
3394

Z
zhangjinchao01 已提交
3395 3396 3397 3398 3399
@config_func
def default_device(val):
    global g_default_device
    g_default_device = val

Q
qijun 已提交
3400

Z
zhangjinchao01 已提交
3401 3402 3403 3404 3405
@config_func
def default_update_hooks(val):
    global g_default_update_hooks
    g_default_update_hooks = val

Q
qijun 已提交
3406

Z
zhangjinchao01 已提交
3407 3408 3409 3410 3411
@config_func
def default_compact_func(val):
    global g_default_compact_func
    g_default_compact_func = val

Q
qijun 已提交
3412

Z
zhangjinchao01 已提交
3413 3414 3415 3416 3417
def make_importer(config_dir, config_args):
    def Import(config_file, local_args={}):
        if not config_file.startswith('/'):
            config_file = config_dir + '/' + config_file
            g_config.config_files.append(config_file)
Q
qijun 已提交
3418 3419 3420
        execfile(config_file,
                 make_config_environment(config_file, config_args), local_args)

Z
zhangjinchao01 已提交
3421 3422
    return Import

Q
qijun 已提交
3423

X
xuwei06 已提交
3424
DEFAULT_SETTING = dict(
Z
zhangjinchao01 已提交
3425 3426 3427 3428 3429
    batch_size=None,
    mini_batch_size=None,
    algorithm='async_sgd',
    async_lagged_grad_discard_ratio=1.5,
    learning_method='momentum',
3430
    gradient_clipping_threshold=None,
Z
zhangjinchao01 已提交
3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452
    num_batches_per_send_parameter=None,
    num_batches_per_get_parameter=None,
    center_parameter_update_method=None,
    learning_rate=1.,
    learning_rate_decay_a=0.,
    learning_rate_decay_b=0.,
    learning_rate_schedule='poly',
    learning_rate_args='',
    l1weight=0.1,
    l2weight=0.,
    l2weight_zero_iter=0,
    c1=0.0001,
    backoff=0.5,
    owlqn_steps=10,
    max_backoff=5,
    average_window=0,
    do_average_in_cpu=False,
    max_average_window=None,
    ada_epsilon=1e-6,
    ada_rou=0.95,
    delta_add_rate=1.0,
    shrink_parameter_value=0,
Q
qijun 已提交
3453 3454 3455
    adam_beta1=0.9,
    adam_beta2=0.999,
    adam_epsilon=1e-8, )
Z
zhangjinchao01 已提交
3456

X
xuwei06 已提交
3457
settings = copy.deepcopy(DEFAULT_SETTING)
X
xuwei06 已提交
3458

Q
qijun 已提交
3459
settings_deprecated = dict(usage_ratio=1., )
Z
zhangjinchao01 已提交
3460 3461 3462 3463

trainer_settings = dict(
    save_dir="./output/model",
    init_model_path=None,
Q
qijun 已提交
3464 3465
    start_pass=0, )

Z
zhangjinchao01 已提交
3466 3467 3468 3469 3470

@config_func
def Settings(**args):
    for k, v in args.iteritems():
        if k == "usage_ratio":
Q
qijun 已提交
3471 3472
            logger.warning(
                "Deprecated: define usage_ratio in DataConfig instead")
Z
zhangjinchao01 已提交
3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483
            if g_config.HasField("data_config"):
                g_config.data_config.__setattr__(k, v)
            settings_deprecated[k] = v
            continue
        elif k in settings:
            settings[k] = v
        elif k in trainer_settings:
            trainer_settings[k] = v
        else:
            logger.fatal('Unkown setting: %s' % k)

Q
qijun 已提交
3484

Z
zhangjinchao01 已提交
3485 3486 3487 3488
@config_func
def cluster_config(**args):
    pass

Q
qijun 已提交
3489

Z
zhangjinchao01 已提交
3490 3491 3492 3493 3494 3495 3496 3497 3498
@config_func
def EnableSubmodelSuffix(flag=True):
    """
    If enabled, the layer and evaluator names in submodel will be automatically
    appended with @submodel_name
    """
    global g_add_submodel_suffix
    g_add_submodel_suffix = flag

Q
qijun 已提交
3499

Z
zhangjinchao01 已提交
3500 3501 3502 3503
def make_config_environment(config_file, config_args):
    def make_setter(k):
        def setter(v):
            logger.fatal("Obsolete: use Settings(%s=%s, ...) instead" % (k, v))
Q
qijun 已提交
3504

Z
zhangjinchao01 已提交
3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519
        return setter

    funcs = {}
    funcs.update(g_config_funcs)

    for k in settings.iterkeys():
        funcs[k] = make_setter(k)
    for k in settings_deprecated.iterkeys():
        funcs[k] = make_setter(k)
    config_dir = os.path.dirname(config_file)
    if not config_dir:
        config_dir = '.'

    funcs.update(
        Import=make_importer(config_dir, config_args),
Q
qijun 已提交
3520
        get_config_arg=make_get_config_arg(config_args), )
Z
zhangjinchao01 已提交
3521 3522 3523 3524 3525

    funcs.update(g_extended_config_funcs)

    return funcs

Q
qijun 已提交
3526

Z
zhangjinchao01 已提交
3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542
def make_get_config_arg(config_args):
    def get_config_arg(name, type, default=None):
        if type == bool:
            s = config_args.get(name)
            if not s:
                return default
            if s == 'True' or s == '1' or s == 'true':
                return True
            if s == 'False' or s == '0' or s == 'false':
                return False
            raise ValueError('Value of config_arg %s is not boolean' % name)
        else:
            return type(config_args.get(name, default))

    return get_config_arg

Q
qijun 已提交
3543

Z
zhangjinchao01 已提交
3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555
def importlib(name):
    __import__(name)
    return sys.modules[name]


def find_caller():
    stack = traceback.extract_stack()
    for s in stack[-4::-1]:
        if not s[0].endswith('config_parser.py'):
            return s[0], s[1], s[2]
    return "(unknown file)", 0, "(unknown function)"

Q
qijun 已提交
3556

Z
zhangjinchao01 已提交
3557 3558 3559 3560
def my_fatal(s):
    logger.critical(s)
    raise Exception()

Y
Yu Yang 已提交
3561

3562
_parse_config_hooks = set()
Y
Yu Yang 已提交
3563 3564


3565 3566 3567 3568 3569 3570 3571
def register_parse_config_hook(f):
    """
    Register a hook function for parse_config. parse_config will invoke the hook
    at the beginning of parse. This make it possible to reset global state for
    for constructing the model.
    """
    _parse_config_hooks.add(f)
Q
qijun 已提交
3572

Y
Yu Yang 已提交
3573

3574
def update_g_config():
Z
zhangjinchao01 已提交
3575
    '''
3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598
    Update g_config after execute config_file or config_functions.
    '''
    for k, v in settings.iteritems():
        if v is None:
            continue
        g_config.opt_config.__setattr__(k, v)

    for k, v in trainer_settings.iteritems():
        if v is None:
            continue
        g_config.__setattr__(k, v)

    for name in g_config.model_config.input_layer_names:
        assert name in g_layer_map, \
            'input name "%s" does not correspond to a layer name' % name
        assert (g_layer_map[name].type == "data" or g_layer_map[name].type == "data_trim"), \
            'The type of input layer "%s" is not "data"' % name
    for name in g_config.model_config.output_layer_names:
        assert name in g_layer_map, \
            'input name "%s" does not correspond to a layer name' % name
    return g_config


3599
def begin_parse():
Z
zhangjinchao01 已提交
3600
    init_config_environment()
3601 3602
    for hook in _parse_config_hooks:
        hook()
Z
zhangjinchao01 已提交
3603 3604 3605 3606 3607

    logger.findCaller = find_caller
    logger.fatal = my_fatal

    g_config.model_config.type = "nn"
X
xuwei06 已提交
3608 3609 3610 3611 3612 3613 3614 3615 3616

    global g_current_submodel, g_root_submodel
    g_root_submodel = g_config.model_config.sub_models.add()
    g_root_submodel.name = 'root'
    g_root_submodel.is_recurrent_layer_group = False
    g_current_submodel = g_root_submodel


def parse_config(trainer_config, config_arg_str):
3617 3618 3619 3620
    '''
    @param config_arg_str: a string of the form var1=val1,var2=val2. It will be
    passed to config script as a dictionary CONFIG_ARGS
    '''
X
xuwei06 已提交
3621

3622
    begin_parse()
X
xuwei06 已提交
3623 3624
    config_args = {}

Z
zhangjinchao01 已提交
3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636
    if config_arg_str:
        config_args = dict([f.split('=') for f in config_arg_str.split(',')])

    global g_command_config_args
    g_command_config_args.update(config_args)

    extension_module_name = config_args.get('extension_module_name')
    if extension_module_name:
        global g_extended_config_funcs
        extension_module = importlib(extension_module_name)
        g_extended_config_funcs = extension_module.get_config_funcs(g_config)

3637 3638
    if hasattr(trainer_config, '__call__'):
        trainer_config.func_globals.update(
L
Luo Tao 已提交
3639
            make_config_environment("", config_args))
3640
        trainer_config()
H
hanchao 已提交
3641
    else:
3642 3643
        execfile(trainer_config,
                 make_config_environment(trainer_config, config_args))
Z
zhangjinchao01 已提交
3644

3645
    return update_g_config()
Z
zhangjinchao01 已提交
3646 3647


3648
def parse_config_and_serialize(trainer_config, config_arg_str):
Z
zhangjinchao01 已提交
3649
    try:
3650
        config = parse_config(trainer_config, config_arg_str)
Z
zhangjinchao01 已提交
3651 3652 3653 3654 3655 3656
        #logger.info(config)
        return config.SerializeToString()
    except:
        traceback.print_exc()
        raise

Q
qijun 已提交
3657

Z
zhangjinchao01 已提交
3658 3659 3660 3661 3662 3663 3664 3665
if __name__ == '__main__':
    try:
        config = parse_config(sys.argv[1], '')
        config.SerializeToString()
        __real_print__(str(config))
    except:
        traceback.print_exc()
        raise