config_parser.py 123.7 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 99 100 101 102
#
# 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
103
    format='[%(levelname)s %(asctime)s %(filename)s:%(lineno)s] %(message)s', )
Z
zhangjinchao01 已提交
104 105 106
logger = logging.getLogger('paddle')
logger.setLevel(logging.INFO)
__real_print__ = print
Q
qijun 已提交
107
print = logger.info
Z
zhangjinchao01 已提交
108 109 110 111

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

Q
qijun 已提交
112

Z
zhangjinchao01 已提交
113 114 115
# 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
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
        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={},
        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=[],
L
Luo Tao 已提交
141 142 143
        g_add_submodel_suffix=False,

        # Whether current layer needs to pass the image height and width.
144 145 146
        # Default value is true, but if it encounters recurrent_layer_group,
        # it will be false. The reason is that image is converted to be sequence,
        # image height will be sequence length, and image width will be feature
L
Luo Tao 已提交
147 148
        # length of each timestep.
        g_pass_height_width=True, ):
Z
zhangjinchao01 已提交
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164

    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
165

Z
zhangjinchao01 已提交
166 167
g_config_funcs = {}

Q
qijun 已提交
168

Z
zhangjinchao01 已提交
169 170 171 172 173
# 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
174

Z
zhangjinchao01 已提交
175 176 177 178 179
# 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
180

Z
zhangjinchao01 已提交
181 182 183 184 185 186
# 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
187

Z
zhangjinchao01 已提交
188 189
    return wrap

Q
qijun 已提交
190

Z
zhangjinchao01 已提交
191 192 193
def gen_parameter_name(layer_name, input_index):
    return '_%s.w%d' % (layer_name, input_index)

Q
qijun 已提交
194

Z
zhangjinchao01 已提交
195 196 197
def gen_bias_parameter_name(layer_name):
    return '_%s.wbias' % layer_name

Q
qijun 已提交
198

Z
zhangjinchao01 已提交
199 200 201
def default(x, default_value):
    return default_value if x is None else x

Q
qijun 已提交
202

Z
zhangjinchao01 已提交
203 204 205 206 207 208
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
209

Z
zhangjinchao01 已提交
210 211
# functions available in config file

Q
qijun 已提交
212

Z
zhangjinchao01 已提交
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
# 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
231

232 233
@config_func
def HasInputsSet():
234
    return len(g_current_submodel.input_layer_names) != 0
235

Z
zhangjinchao01 已提交
236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259

# 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
260
    name = MakeLayerNameInParentSubmodel(name)  #rename in nested submodel
Z
zhangjinchao01 已提交
261 262 263 264 265 266 267 268 269

    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
270

Z
zhangjinchao01 已提交
271
@config_func
Q
qijun 已提交
272
def SubModelEnd(name=None):
Z
zhangjinchao01 已提交
273
    global g_current_submodel, g_root_submodel, g_submodel_stack
Q
qijun 已提交
274 275
    config_assert(g_current_submodel is not g_root_submodel,
                  "submodel not begin")
Z
zhangjinchao01 已提交
276
    if name is not None:
Q
qijun 已提交
277 278 279
        config_assert(
            g_current_submodel.name == MakeLayerNameInParentSubmodel(name),
            "submodel name error")
Z
zhangjinchao01 已提交
280 281 282

    g_current_submodel = g_submodel_stack.pop()

Q
qijun 已提交
283

Z
zhangjinchao01 已提交
284 285
def MakeLayerNameInParentSubmodel(name):
    suffix = ""
286 287
    if len(g_submodel_stack) > 1:
        suffix = "@" + g_submodel_stack[-1].name
Z
zhangjinchao01 已提交
288 289
    return name + suffix

Q
qijun 已提交
290

Z
zhangjinchao01 已提交
291 292 293
def GetLayerBaseName(name):
    return name.split('@')[0]

Q
qijun 已提交
294 295

def MakeLayerNameInSubmodel(name, submodel_name=None):
Z
zhangjinchao01 已提交
296 297
    global g_current_submodel
    global g_add_submodel_suffix
Q
qijun 已提交
298 299
    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
300 301 302 303 304
        return name
    if submodel_name is None:
        submodel_name = g_current_submodel.name
    return name + "@" + submodel_name

Q
qijun 已提交
305

Z
zhangjinchao01 已提交
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328
# 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,
329 330
                                            seq_reversed=False,
                                            target_inlinkname=""):
Z
zhangjinchao01 已提交
331 332 333 334 335 336 337
    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
338
    g_current_submodel.target_inlinkid = -1
Z
zhangjinchao01 已提交
339
    in_links_count = 0
340
    for linkid, link in enumerate(in_links):
Z
zhangjinchao01 已提交
341 342 343 344 345 346
        if isinstance(link, basestring):
            name = link
            has_subseq = False
        else:
            name = link.link_name
            has_subseq = link.has_subseq
347 348 349 350
        # assign target_inlinkid according to target_inlinkname
        if target_inlinkname == name:
            g_current_submodel.target_inlinkid = linkid

Z
zhangjinchao01 已提交
351 352 353
        if in_links_count == 0:
            in_links_has_subseq = has_subseq
        else:
Q
qijun 已提交
354 355 356 357
            config_assert(
                in_links_has_subseq == has_subseq,
                "The sequence type of in_links should be the same in RecurrentLayerGroup"
            )
Z
zhangjinchao01 已提交
358 359 360 361 362 363 364
        in_links_count += 1
        layer_name = MakeLayerNameInParentSubmodel(name)
        layer = g_layer_map[layer_name]
        if has_subseq:
            SequenceScatterAgentLayer(name=name, size=layer.size)
        else:
            ScatterAgentLayer(name=name, size=layer.size)
365

Z
zhangjinchao01 已提交
366 367 368 369 370
        pair = g_current_submodel.in_links.add()
        pair.layer_name = layer_name
        pair.link_name = MakeLayerNameInSubmodel(name)
        pair.has_subseq = has_subseq

Q
qijun 已提交
371

Z
zhangjinchao01 已提交
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
@config_func
def RecurrentLayerGroupSetOutLink(link):
    if isinstance(link, basestring):
        name = link
        has_subseq = False
    else:
        name = link.link_name
        has_subseq = link.has_subseq
    layer_name = MakeLayerNameInParentSubmodel(name)
    pair = g_current_submodel.out_links.add()
    pair.layer_name = MakeLayerNameInSubmodel(name)
    pair.link_name = layer_name
    pair.has_subseq = has_subseq


def RecurrentLayerGroupSetGenerator(generator=None):
Q
qijun 已提交
388
    generator.eos_layer_name = MakeLayerNameInSubmodel(generator.eos_layer_name)
Z
zhangjinchao01 已提交
389 390 391 392 393 394 395 396
    g_current_submodel.generator.CopyFrom(generator)


@config_func
def RecurrentLayerGroupBegin(name,
                             in_links,
                             out_links,
                             generator=None,
397
                             target_inlinkname="",
Z
zhangjinchao01 已提交
398
                             seq_reversed=False):
Q
qijun 已提交
399
    RecurrentLayerGroupWithoutOutLinksBegin(name, in_links, seq_reversed,
400
                                            target_inlinkname)
Z
zhangjinchao01 已提交
401 402 403 404 405
    for link in out_links:
        RecurrentLayerGroupSetOutLink(link)

    if generator is not None:
        RecurrentLayerGroupSetGenerator(generator)
Q
qijun 已提交
406 407 408 409 410
        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
411 412 413 414 415 416 417


@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
418
    for pair in g_current_submodel.memories:  #check exist
Z
zhangjinchao01 已提交
419
        layer = g_layer_map[pair.layer_name]
Y
Yu Yang 已提交
420 421
        config_assert(layer is not None,
                      "memory declare wrong name:%s" % pair.layer_name)
Z
zhangjinchao01 已提交
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
        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)
        elif pair.has_subseq:
            SequenceGatherAgentLayer(name=agent_name, size=layer.size)
        else:
            GatherAgentLayer(name=agent_name, size=layer.size)

Q
qijun 已提交
440

Z
zhangjinchao01 已提交
441 442 443 444 445 446
# 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
447

Z
zhangjinchao01 已提交
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464
@config_class
class Bias(Cfg):
    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,
Q
qijun 已提交
465
            is_shared=None, ):
Z
zhangjinchao01 已提交
466 467
        self.add_keys(locals())

Q
qijun 已提交
468

Z
zhangjinchao01 已提交
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488
# Define one input for a layer
@config_class
class Input(Cfg):
    def __init__(
            self,
            input_layer_name,
            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,
            sparse_update=None,
            gradient_clipping_threshold=None,
            conv=None,
L
liaogang 已提交
489
            bilinear_interp=None,
Z
zhangjinchao01 已提交
490 491 492 493
            norm=None,
            pool=None,
            image=None,
            block_expand=None,
494
            maxout=None,
Q
qijun 已提交
495
            spp=None,
Z
zhangjinchao01 已提交
496 497 498 499 500
            format=None,
            nnz=None,
            is_static=None,
            is_shared=None,
            update_hooks=None,
501
            input_layer_argument=None,
D
dangqingqing 已提交
502 503 504 505 506
            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
507
        self.add_keys(locals())
D
dangqingqing 已提交
508 509 510
        self.input_layer_name = MakeLayerNameInSubmodel(
            input_layer_name
        ) if make_layer_name_in_submodel else input_layer_name
Z
zhangjinchao01 已提交
511

Q
qijun 已提交
512

Z
zhangjinchao01 已提交
513 514 515
# Define a projection for iexed layer
@config_class
class Projection(Input):
Q
qijun 已提交
516 517
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
518 519 520
    def __init__(
            self,
            input_layer_name,
Q
qijun 已提交
521
            size=0,  # projection output size
Z
zhangjinchao01 已提交
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
            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,
            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
541
            input_layer_argument=None, ):
Z
zhangjinchao01 已提交
542 543 544 545 546 547 548 549 550 551 552 553 554
        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
555

Z
zhangjinchao01 已提交
556 557
    def calc_parameter_size(self, input_size, output_size):
        raise NotimplementedError
Q
qijun 已提交
558

Z
zhangjinchao01 已提交
559 560 561 562 563 564 565 566 567 568
    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
569

Z
zhangjinchao01 已提交
570 571
    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
572

Z
zhangjinchao01 已提交
573 574 575
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
576

Z
zhangjinchao01 已提交
577 578 579 580 581 582
# 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
583 584 585
    def __init__(self, input_layer_name, offset, **xargs):
        super(IdentityOffsetProjection, self).__init__(input_layer_name,
                                                       **xargs)
Z
zhangjinchao01 已提交
586 587 588 589
        self.proj_conf.offset = offset

    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
590

Z
zhangjinchao01 已提交
591 592 593
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
594

Z
zhangjinchao01 已提交
595 596 597 598 599 600 601
# 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
602

Z
zhangjinchao01 已提交
603 604
    def calc_parameter_size(self, input_size, output_size):
        return output_size
Q
qijun 已提交
605

Z
zhangjinchao01 已提交
606 607 608
    def calc_parameter_dims(self, input_size, output_size):
        return [1, output_size]

L
Luo Tao 已提交
609

X
xuwei06 已提交
610 611 612 613 614 615 616 617 618 619 620 621 622 623
# 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
624

Z
zhangjinchao01 已提交
625 626 627 628 629 630
@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
631

Z
zhangjinchao01 已提交
632 633 634
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
635

Z
zhangjinchao01 已提交
636 637 638 639 640 641
@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
642

Z
zhangjinchao01 已提交
643 644 645
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
646

Z
zhangjinchao01 已提交
647 648 649 650 651 652
@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
653

Z
zhangjinchao01 已提交
654 655 656
    def calc_parameter_dims(self, input_size, output_size):
        return [output_size, input_size]

Q
qijun 已提交
657

Z
zhangjinchao01 已提交
658 659 660 661
@config_class
class ContextProjection(Projection):
    type = 'context'

Q
qijun 已提交
662 663
    def __init__(self, input_layer_name, context_start, context_length,
                 trainable_padding, **xargs):
Z
zhangjinchao01 已提交
664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686
        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


687 688 689 690
@config_class
class ConvProjection(Projection):
    type = 'conv'

Q
qijun 已提交
691 692 693 694 695
    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
696 697 698 699 700
        super(ConvProjection, self).__init__(input_layer_name, **xargs)

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

Q
qijun 已提交
701
        parse_conv(conv_conf, input_layer_name, self.proj_conf.conv_conf,
702
                   num_filters)
L
Luo Tao 已提交
703 704 705
        self.proj_conf.output_size = self.proj_conf.conv_conf.output_x * \
                                     self.proj_conf.conv_conf.output_y * \
                                     num_filters
706 707 708 709 710 711 712 713 714

    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
715 716
        gr = self.proj_conf.conv_conf.groups
        return co * ci * fh * fw / gr
717 718 719 720 721 722 723

    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
724

Z
zhangjinchao01 已提交
725 726 727
# Define a operator for mixed layer
@config_class
class Operator(Cfg):
Q
qijun 已提交
728 729
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
730 731
    def __init__(
            self,
Q
qijun 已提交
732
            input_layer_names, ):
Z
zhangjinchao01 已提交
733 734 735 736 737 738 739 740 741 742
        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
743

Z
zhangjinchao01 已提交
744 745 746
@config_class
class DotMulOperator(Operator):
    type = 'dot_mul'
Q
qijun 已提交
747 748 749

    def __init__(self, input_layer_names, scale=None, **xargs):
        super(DotMulOperator, self).__init__(input_layer_names, **xargs)
Z
zhangjinchao01 已提交
750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767
        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
768 769 770 771 772 773 774

    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
775 776 777
        if num_filters is not None:
            self.operator_conf.num_filters = num_filters

778 779
        parse_conv(conv_conf,
                   MakeLayerNameInSubmodel(input_layer_names[0]),
Q
qijun 已提交
780
                   self.operator_conf.conv_conf, num_filters)
L
Luo Tao 已提交
781 782 783
        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
784 785 786

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

787 788
    def calc_output_size(self, input_sizes):
        return self.operator_conf.output_size
Z
zhangjinchao01 已提交
789 790 791 792 793


# please refer to the comments in proto/ModelConfig.proto
@config_class
class Conv(Cfg):
Q
qijun 已提交
794 795 796 797 798 799 800 801 802 803 804 805 806
    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
807 808
        self.add_keys(locals())
        if filter_size_y is None:
Q
qijun 已提交
809
            self.filter_size_y = filter_size
Z
zhangjinchao01 已提交
810
        if padding_y is None:
Q
qijun 已提交
811
            self.padding_y = padding
Z
zhangjinchao01 已提交
812
        if stride_y is None:
Q
qijun 已提交
813
            self.stride_y = stride
Z
zhangjinchao01 已提交
814
        if output_x is not None:
Q
qijun 已提交
815 816
            config_assert(output_x <= 0)

Z
zhangjinchao01 已提交
817

L
liaogang 已提交
818 819
@config_class
class BilinearInterp(Cfg):
L
Luo Tao 已提交
820
    def __init__(self, out_size_x=None, out_size_y=None, channels=None):
L
liaogang 已提交
821 822
        self.add_keys(locals())

Q
qijun 已提交
823

Z
zhangjinchao01 已提交
824 825
@config_class
class Pool(Cfg):
D
dangqingqing 已提交
826 827 828 829 830 831 832 833 834 835 836 837
    def __init__(
            self,
            pool_type,
            channels,
            size_x,
            size_y=None,
            img_width=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
838
        self.add_keys(locals())
Q
qijun 已提交
839 840


Q
qijun 已提交
841
@config_class
Q
qijun 已提交
842
class SpatialPyramidPool(Cfg):
L
Luo Tao 已提交
843
    def __init__(self, pool_type, pyramid_height, channels):
Q
qijun 已提交
844
        self.add_keys(locals())
Z
zhangjinchao01 已提交
845

Q
qijun 已提交
846

Z
zhangjinchao01 已提交
847 848
@config_class
class Norm(Cfg):
Q
qijun 已提交
849 850 851 852 853 854 855 856 857
    def __init__(self,
                 norm_type,
                 channels,
                 size,
                 scale,
                 pow,
                 output_x=None,
                 img_size=None,
                 blocked=None):
Z
zhangjinchao01 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 Image(Cfg):
Q
qijun 已提交
863
    def __init__(self, channels, img_size=None):
Z
zhangjinchao01 已提交
864 865
        self.add_keys(locals())

Q
qijun 已提交
866

Z
zhangjinchao01 已提交
867 868
@config_class
class BlockExpand(Cfg):
Q
qijun 已提交
869 870 871 872 873 874 875 876 877 878 879 880
    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
881 882
        self.add_keys(locals())

Q
qijun 已提交
883

884 885
@config_class
class MaxOut(Cfg):
Q
qijun 已提交
886
    def __init__(self, channels, groups, img_size_x=0, img_size_y=0):
887 888
        self.add_keys(locals())

Q
qijun 已提交
889

Z
zhangjinchao01 已提交
890 891 892 893 894 895 896 897 898 899 900 901 902
def DataBase(async_load_data=False,
             constant_slots=None,
             data_ratio=1,
             is_main_data=True,
             usage_ratio=None):
    # 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
903 904
    data_config.data_ratio = data_ratio
    data_config.is_main_data = is_main_data
Z
zhangjinchao01 已提交
905

Q
qijun 已提交
906
    usage_ratio = default(usage_ratio, settings_deprecated["usage_ratio"])
Z
zhangjinchao01 已提交
907 908 909 910 911 912
    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
913

Z
zhangjinchao01 已提交
914
@config_func
Q
qijun 已提交
915 916 917 918 919
def SimpleData(files=None,
               feat_dim=None,
               context_len=None,
               buffer_capacity=None,
               **xargs):
Z
zhangjinchao01 已提交
920 921 922 923 924 925 926 927 928 929
    data_config = DataBase(**xargs)
    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
930

Z
zhangjinchao01 已提交
931
@config_func
Q
qijun 已提交
932 933 934 935 936 937 938 939 940 941
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):
Z
zhangjinchao01 已提交
942 943 944
    data_config = DataBase(**xargs)
    data_config.type = 'py'
    if load_data_module in g_py_module_name_list:
Q
qijun 已提交
945

Z
zhangjinchao01 已提交
946 947 948
        def get_path(module):
            m = __import__(load_data_module)
            return os.path.split(os.path.realpath(m.__file__))[0]
Q
qijun 已提交
949

Z
zhangjinchao01 已提交
950 951 952
        # 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
953 954
        module_new_name = "%s_copy_%d" % (load_data_module,
                                          len(g_py_module_name_list))
Z
zhangjinchao01 已提交
955
        g_py_module_name_list.append(module_new_name)
Q
qijun 已提交
956 957 958 959
        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
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
        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
984

Z
zhangjinchao01 已提交
985
@config_func
Q
qijun 已提交
986 987 988 989 990 991 992
def ProtoData(files=None,
              type=None,
              file_group_queue_capacity=None,
              load_file_count=None,
              constant_slots=None,
              load_thread_num=None,
              **xargs):
Z
zhangjinchao01 已提交
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
    data_config = DataBase(**xargs)
    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
1013

Z
zhangjinchao01 已提交
1014 1015
#real data for training is actually provided by "sub_data" data providers.
@config_func
Q
qijun 已提交
1016
def MultiData(sub_data=[]):
Z
zhangjinchao01 已提交
1017 1018 1019 1020 1021
    data_config = DataConfig()
    data_config.type = 'multi'
    data_config.sub_data_configs.extend(sub_data)
    return data_config

Q
qijun 已提交
1022

Z
zhangjinchao01 已提交
1023
@config_func
Q
qijun 已提交
1024 1025 1026 1027 1028 1029 1030
def Data(type,
         files=None,
         feat_dim=None,
         slot_dims=None,
         context_len=None,
         buffer_capacity=None,
         **xargs):
Z
zhangjinchao01 已提交
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065

    data_config = DataBase(**xargs)
    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
1066

L
Luo Tao 已提交
1067 1068
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1069 1070 1071 1072 1073 1074 1075
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
1076

1077
#calcualte image_size based on output_size for de-convolution (ConvTransLayer).
L
Luo Tao 已提交
1078
#It is the reverse function of cnn_output_size
1079
def cnn_image_size(output_size, filter_size, padding, stride, caffe_mode):
L
Luo Tao 已提交
1080 1081 1082
    img_size = (output_size - 1) * stride + filter_size - 2 * padding
    if not caffe_mode:
        img_size = img_size + 1
1083 1084
    return img_size

Q
qijun 已提交
1085

L
Luo Tao 已提交
1086
def get_img_size(input_layer_name, channels):
L
Luo Tao 已提交
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
    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


Z
zhangjinchao01 已提交
1105 1106
def parse_pool(pool, input_layer_name, pool_conf):
    pool_conf.pool_type = pool.pool_type
Q
qijun 已提交
1107 1108 1109
    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
1110
                  "['max-projection', 'avg-projection', "
Q
qijun 已提交
1111
                  "'cudnn-max-pool', 'cudnn-avg-pool']" % pool.pool_type)
Z
zhangjinchao01 已提交
1112 1113 1114 1115 1116 1117

    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
1118
    pool_conf.stride_y = default(pool.stride_y, pool_conf.stride)
Z
zhangjinchao01 已提交
1119

L
Luo Tao 已提交
1120
    pool_conf.img_size, pool_conf.img_size_y = \
L
Luo Tao 已提交
1121
        get_img_size(input_layer_name, pool.channels)
Z
zhangjinchao01 已提交
1122

1123
    config_assert(not pool.start, "start is deprecated in pooling.")
Z
zhangjinchao01 已提交
1124

1125
    if pool.padding is not None:
Z
zhangjinchao01 已提交
1126
        pool_conf.padding = pool.padding
1127
    pool_conf.padding_y = default(pool.padding_y, pool_conf.padding)
D
dangqingqing 已提交
1128 1129 1130 1131 1132 1133
    pool_conf.output_x = cnn_output_size(pool_conf.img_size, pool_conf.size_x,
                                         pool_conf.padding, pool_conf.stride,
                                         False)
    pool_conf.output_y = cnn_output_size(pool_conf.img_size_y, pool_conf.size_y,
                                         pool_conf.padding_y,
                                         pool_conf.stride_y, False)
Q
qijun 已提交
1134

Z
zhangjinchao01 已提交
1135

Q
qijun 已提交
1136
def parse_spp(spp, input_layer_name, spp_conf):
L
Luo Tao 已提交
1137
    parse_image(spp, input_layer_name, spp_conf.image_conf)
Q
qijun 已提交
1138 1139
    spp_conf.pool_type = spp.pool_type
    config_assert(spp.pool_type in ['max-projection', 'avg-projection'],
Q
qijun 已提交
1140 1141
                  "pool-type %s is not in "
                  "['max-projection', 'avg-projection']" % spp.pool_type)
Q
qijun 已提交
1142
    spp_conf.pyramid_height = spp.pyramid_height
Q
qijun 已提交
1143

Q
qijun 已提交
1144

Z
zhangjinchao01 已提交
1145 1146
def parse_image(image, input_layer_name, image_conf):
    image_conf.channels = image.channels
L
Luo Tao 已提交
1147
    image_conf.img_size, image_conf.img_size_y = \
L
Luo Tao 已提交
1148
        get_img_size(input_layer_name, image_conf.channels)
Q
qijun 已提交
1149

Z
zhangjinchao01 已提交
1150 1151 1152 1153

def parse_norm(norm, input_layer_name, norm_conf):
    norm_conf.norm_type = norm.norm_type
    config_assert(norm.norm_type in ['rnorm', 'cmrnorm-projection'],
Q
qijun 已提交
1154 1155
                  "norm-type %s is not in [rnorm, 'cmrnorm-projection']" %
                  norm.norm_type)
Z
zhangjinchao01 已提交
1156 1157 1158 1159 1160 1161
    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
1162
    norm_conf.img_size, norm_conf.img_size_y = \
L
Luo Tao 已提交
1163
        get_img_size(input_layer_name, norm.channels)
Z
zhangjinchao01 已提交
1164
    norm_conf.output_x = norm_conf.img_size
L
Luo Tao 已提交
1165
    norm_conf.output_y = norm_conf.img_size_y
Z
zhangjinchao01 已提交
1166 1167 1168
    if norm.norm_type in ['cmrnorm-projection']:
        norm_conf.scale /= norm.size
    else:
Q
qijun 已提交
1169 1170
        norm_conf.scale /= norm.size**2

1171

L
Luo Tao 已提交
1172 1173
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1174
def parse_conv(conv, input_layer_name, conv_conf, num_filters, trans=False):
Z
zhangjinchao01 已提交
1175 1176 1177 1178 1179 1180 1181 1182 1183
    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
1184

1185
    if not trans:
1186
        conv_conf.filter_channels = conv.channels / conv.groups
L
Luo Tao 已提交
1187
        conv_conf.img_size, conv_conf.img_size_y = \
L
Luo Tao 已提交
1188
            get_img_size(input_layer_name, conv.channels)
1189
        conv_conf.output_x = cnn_output_size(
Q
qijun 已提交
1190 1191
            conv_conf.img_size, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1192 1193 1194
        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)
1195
    else:
1196
        conv_conf.filter_channels = num_filters / conv.groups
L
Luo Tao 已提交
1197
        conv_conf.output_x, conv_conf.output_y = \
L
Luo Tao 已提交
1198
            get_img_size(input_layer_name, conv.channels)
1199
        conv_conf.img_size = cnn_image_size(
Q
qijun 已提交
1200 1201
            conv_conf.output_x, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1202
        conv_conf.img_size_y = cnn_image_size(
L
Luo Tao 已提交
1203 1204
            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
1205

1206

Z
zhangjinchao01 已提交
1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219
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:
1220
        block_expand_conf.output_x = cnn_output_size(
1221
            block_expand.img_size_x, block_expand.block_x,
1222
            block_expand.padding_x, block_expand.stride_x, False)
Z
zhangjinchao01 已提交
1223 1224

    if block_expand_conf.img_size_y == 0:
1225
        block_expand_conf.output_y = 0
Z
zhangjinchao01 已提交
1226
    else:
1227
        block_expand_conf.output_y = cnn_output_size(
1228
            block_expand.img_size_y, block_expand.block_y,
1229
            block_expand.padding_y, block_expand.stride_y, False)
Z
zhangjinchao01 已提交
1230

Q
qijun 已提交
1231

1232
def parse_maxout(maxout, input_layer_name, maxout_conf):
L
Luo Tao 已提交
1233
    parse_image(maxout, input_layer_name, maxout_conf.image_conf)
1234
    maxout_conf.groups = maxout.groups
1235

Q
qijun 已提交
1236

Z
zhangjinchao01 已提交
1237 1238 1239 1240 1241 1242
# Define an evaluator
@config_func
def Evaluator(
        name,
        type,
        inputs,
Q
qijun 已提交
1243 1244 1245 1246 1247 1248 1249
        chunk_scheme=None,
        num_chunk_types=None,
        classification_threshold=None,
        positive_label=None,
        dict_file=None,
        result_file=None,
        num_results=None,
1250 1251
        delimited=None,
        excluded_chunk_types=None, ):
Z
zhangjinchao01 已提交
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
    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)

1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
    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
    if delimited is not None:
        evaluator.delimited = delimited
Z
zhangjinchao01 已提交
1279

1280 1281 1282
    if excluded_chunk_types:
        evaluator.excluded_chunk_types.extend(excluded_chunk_types)

Q
qijun 已提交
1283

Z
zhangjinchao01 已提交
1284 1285 1286 1287 1288
class LayerBase(object):
    def __init__(
            self,
            name,
            type,
Q
qijun 已提交
1289
            size,  # size can be 0. In this case, subclass should set it.
Z
zhangjinchao01 已提交
1290 1291 1292 1293
            inputs,
            device=None,
            active_type="",
            drop_rate=0.,
1294
            coeff=None):
Z
zhangjinchao01 已提交
1295
        config_assert('@' not in name,
Q
qijun 已提交
1296
                      "layer name: %s contain special character @" % name)
Z
zhangjinchao01 已提交
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311
        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()
1312
        assert isinstance(self.config, LayerConfig)
Z
zhangjinchao01 已提交
1313 1314 1315
        self.config.name = name
        self.config.type = type
        self.config.active_type = active_type
1316 1317
        if coeff is not None:
            self.config.coeff = float(coeff)
Z
zhangjinchao01 已提交
1318 1319 1320 1321 1322 1323 1324
        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
1325
        elif g_default_device is not None:
Z
zhangjinchao01 已提交
1326 1327 1328 1329 1330 1331 1332 1333 1334
            self.config.device = g_default_device

        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
1335 1336
                    input_layer_name=input,
                    parameter_name=gen_parameter_name(name, input_index))
Z
zhangjinchao01 已提交
1337 1338 1339 1340 1341 1342 1343 1344
                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
1345
                self.operators.append(input)
Z
zhangjinchao01 已提交
1346 1347 1348 1349
                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
1350
                raise ValueError('Wrong type for inputs: %s' % type_of(input))
Z
zhangjinchao01 已提交
1351
            config_assert(input_layer_name in g_layer_map,
Q
qijun 已提交
1352 1353
                          "Unknown input layer '%s' for layer %s" %
                          (input_layer_name, name))
Z
zhangjinchao01 已提交
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364
            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)

L
Luo Tao 已提交
1365 1366 1367 1368 1369 1370
        if self.config.type != 'data' and g_pass_height_width:
            height = self.get_input_layer(0).height
            width = self.get_input_layer(0).width
            if height and width:
                self.set_layer_height_width(height, width)

Z
zhangjinchao01 已提交
1371 1372 1373 1374 1375 1376
    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
1377
            bias,  # True/False or BiasCfg
Z
zhangjinchao01 已提交
1378
            size,
Q
qijun 已提交
1379 1380 1381
            dims=None,
            for_self=True,  # whether create bias for layer self
    ):
Z
zhangjinchao01 已提交
1382 1383 1384 1385 1386 1387

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

Q
qijun 已提交
1388 1389 1390
        config_assert(
            type_of(bias) == bool or type_of(bias) == Bias,
            'Incorrect type for bias: %s' % type_of(bias))
Z
zhangjinchao01 已提交
1391 1392 1393 1394 1395 1396 1397 1398 1399

        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:
1400 1401
                assert isinstance(self.config, LayerConfig)

Z
zhangjinchao01 已提交
1402 1403 1404
                Parameter(
                    bias.parameter_name,
                    size,
Q
qijun 已提交
1405 1406
                    self.config.device
                    if self.config.HasField('device') else None,
Z
zhangjinchao01 已提交
1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
                    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
1418 1419
                    gradient_clipping_threshold=bias.
                    gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1420
                    is_static=bias.is_static,
Q
qijun 已提交
1421
                    is_shared=bias.is_shared, )
Z
zhangjinchao01 已提交
1422 1423 1424 1425 1426
            if for_self:
                self.config.bias_parameter_name = bias.parameter_name
            else:
                return bias.parameter_name

Q
qijun 已提交
1427 1428 1429 1430 1431 1432
    def create_input_parameter(self,
                               input_index,
                               size,
                               dims=None,
                               sparse=None,
                               format=None):
Z
zhangjinchao01 已提交
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
        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
1447 1448
            config_assert(size == para.size, (
                'Shared parameter "%s" does not ' + 'have same size: %s vs. %s')
Z
zhangjinchao01 已提交
1449 1450
                          % (input_config.parameter_name, para.size, size))

Q
qijun 已提交
1451 1452
            config_assert(dims == para.dims, (
                'Shared parameter "%s" does not ' + 'have same dims: %s vs. %s')
Z
zhangjinchao01 已提交
1453 1454 1455 1456 1457 1458
                          % (input_config.parameter_name, para.dims, dims))
            return

        Parameter(
            input_config.parameter_name,
            size,
1459
            self.config.device if self.config.HasField("device") else None,
Z
zhangjinchao01 已提交
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
            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
1472 1473
            gradient_clipping_threshold=input_config.
            gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1474 1475 1476 1477
            sparse=sparse,
            format=format,
            is_static=input_config.is_static,
            is_shared=input_config.is_shared,
Q
qijun 已提交
1478
            update_hooks=input_config.update_hooks)
Z
zhangjinchao01 已提交
1479 1480 1481 1482 1483 1484 1485 1486 1487

    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
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
    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
1505

Z
zhangjinchao01 已提交
1506 1507
@config_layer('multi_class_cross_entropy_with_selfnorm')
class MultiClassCrossEntropySelfNormCostLayer(LayerBase):
Q
qijun 已提交
1508 1509 1510
    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
1511 1512
        self.config.softmax_selfnorm_alpha = softmax_selfnorm_alpha

Q
qijun 已提交
1513

Z
zhangjinchao01 已提交
1514 1515
@config_layer('fc')
class FCLayer(LayerBase):
Q
qijun 已提交
1516
    def __init__(self, name, size, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
        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
1527 1528
            else:
                sparse = None
Z
zhangjinchao01 已提交
1529

Q
qijun 已提交
1530 1531
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1532 1533
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1534

Z
zhangjinchao01 已提交
1535 1536
@config_layer('selective_fc')
class SelectiveFCLayer(LayerBase):
Q
qijun 已提交
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
    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
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
        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
1567 1568
                          ("if indices of selected columns are not specified, "
                           "selective_fc Layer has at least two inputs"))
Z
zhangjinchao01 已提交
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580
            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
1581 1582
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1583 1584
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1585

1586 1587
@config_layer('print')
class PrintLayer(LayerBase):
Q
qijun 已提交
1588
    def __init__(self, name, inputs):
1589 1590
        super(PrintLayer, self).__init__(name, 'print', 0, inputs)

Q
qijun 已提交
1591

Y
yuan 已提交
1592 1593
@config_layer('priorbox')
class PriorBoxLayer(LayerBase):
G
gaoyuan 已提交
1594 1595
    def __init__(self, name, inputs, size, min_size, max_size, aspect_ratio,
                 variance):
Y
yuan 已提交
1596
        super(PriorBoxLayer, self).__init__(name, 'priorbox', 0, inputs)
G
gaoyuan 已提交
1597
        config_assert(len(inputs) == 2, 'PriorBoxLayer must have 2 inputs')
G
gaoyuan 已提交
1598 1599 1600 1601 1602 1603 1604
        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
1605
        config_assert(len(variance) == 4, 'The variance must have 4 inputs')
Y
yuan 已提交
1606 1607 1608 1609 1610 1611
        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
1612

Z
zhangjinchao01 已提交
1613 1614
@config_layer('data')
class DataLayer(LayerBase):
L
Luo Tao 已提交
1615
    def __init__(self, name, size, height=None, width=None, device=None):
Q
qijun 已提交
1616 1617
        super(DataLayer, self).__init__(
            name, 'data', size, inputs=[], device=device)
L
Luo Tao 已提交
1618 1619
        if height and width:
            self.set_layer_height_width(height, width)
Q
qijun 已提交
1620

Z
zhangjinchao01 已提交
1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647

'''
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
1648 1649


Z
zhangjinchao01 已提交
1650 1651
@config_layer('data_norm')
class DataNormLayer(LayerBase):
Q
qijun 已提交
1652
    def __init__(self, name, inputs, data_norm_strategy="z-score", device=None):
Z
zhangjinchao01 已提交
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
        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
1664

Z
zhangjinchao01 已提交
1665 1666 1667
@config_layer('prelu')
class ParameterReluLayer(LayerBase):
    layer_type = 'prelu'
Q
qijun 已提交
1668 1669

    def __init__(self, name, inputs, partial_sum=1, **args):
Z
zhangjinchao01 已提交
1670 1671 1672 1673 1674 1675 1676 1677
        super(ParameterReluLayer, self).__init__(
            name, self.layer_type, 0, inputs=inputs, **args)
        config_assert(len(self.inputs) == 1)
        config_assert(self.input_layer.size % partial_sum == 0)
        input_layer = self.get_input_layer(0)
        self.set_layer_size(input_layer.size)
        self.create_input_parameter(0, input_layer.size / partial_sum)

Q
qijun 已提交
1678

Z
zhangjinchao01 已提交
1679 1680 1681
@config_layer('conv')
class ConvLayerBase(LayerBase):
    layer_type = 'conv'
Q
qijun 已提交
1682 1683 1684 1685 1686 1687 1688 1689

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
Z
zhangjinchao01 已提交
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
        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
1706
            (parallel_nn == 0 or self.config.device > -1)):
Z
zhangjinchao01 已提交
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
            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
1719 1720
            parse_conv(self.inputs[input_index].conv, input_layer.name,
                       conv_conf, num_filters)
Z
zhangjinchao01 已提交
1721 1722
            psize = self.calc_parameter_size(conv_conf)
            self.create_input_parameter(input_index, psize)
L
Luo Tao 已提交
1723 1724
            self.set_cnn_layer(name, conv_conf.output_y, conv_conf.output_x,
                               self.config.num_filters)
Z
zhangjinchao01 已提交
1725 1726 1727 1728 1729 1730 1731 1732 1733 1734

        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
1735

Z
zhangjinchao01 已提交
1736 1737 1738 1739
@config_layer('exconv')
class ConvLayer(ConvLayerBase):
    layer_type = 'exconv'

Q
qijun 已提交
1740

Z
zhangjinchao01 已提交
1741 1742 1743 1744
@config_layer('cudnn_conv')
class ConvLayer(ConvLayerBase):
    layer_type = 'cudnn_conv'

1745 1746 1747 1748

@config_layer('convt')
class ConvTransLayerBase(LayerBase):
    layer_type = 'convt'
Q
qijun 已提交
1749 1750 1751 1752 1753 1754 1755 1756

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
1757
        super(ConvTransLayerBase, self).__init__(
1758 1759 1760 1761 1762 1763 1764 1765
            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))

1766 1767
        # cudnn_convt has not been implemented so use exconvt only
        self.layer_type = "exconvt"
1768 1769 1770 1771 1772 1773 1774 1775
        # 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)
1776
            parse_conv(
1777 1778
                self.inputs[input_index].conv,
                input_layer.name,
1779
                self.config.inputs[input_index].conv_conf,
1780
                num_filters,
1781
                trans=True)
1782 1783 1784 1785 1786
            conv_conf = self.config.inputs[input_index].conv_conf
            psize = self.calc_parameter_size(conv_conf)
            print("output size for %s is %d " % (name, conv_conf.output_x))
            self.create_input_parameter(input_index, psize)
            self.set_layer_size(
Q
qijun 已提交
1787
                (conv_conf.img_size**2) * self.config.num_filters)
1788 1789 1790 1791 1792 1793 1794

        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):
1795
        return conv_conf.channels * conv_conf.filter_channels \
1796 1797
                    * (conv_conf.filter_size * conv_conf.filter_size_y)

Q
qijun 已提交
1798

1799 1800 1801 1802
@config_layer('exconvt')
class ConvTransLayer(ConvTransLayerBase):
    layer_type = 'exconvt'

Q
qijun 已提交
1803

Z
zhangjinchao01 已提交
1804 1805
@config_layer('norm')
class NormLayer(LayerBase):
Q
qijun 已提交
1806 1807 1808
    def __init__(self, name, inputs, device=None):
        super(NormLayer, self).__init__(
            name, 'norm', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1809 1810 1811
        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
1812 1813 1814 1815
            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)
Q
qijun 已提交
1816

Z
zhangjinchao01 已提交
1817 1818 1819

@config_layer('pool')
class PoolLayer(LayerBase):
Q
qijun 已提交
1820 1821 1822
    def __init__(self, name, inputs, device=None):
        super(PoolLayer, self).__init__(
            name, 'pool', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1823 1824 1825
        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
1826 1827 1828 1829
            parse_pool(self.inputs[input_index].pool, input_layer.name,
                       pool_conf)
            self.set_cnn_layer(name, pool_conf.output_y, pool_conf.output_x,
                               pool_conf.channels)
Q
qijun 已提交
1830

Z
zhangjinchao01 已提交
1831

Q
qijun 已提交
1832 1833
@config_layer('spp')
class SpatialPyramidPoolLayer(LayerBase):
Q
qijun 已提交
1834 1835 1836
    def __init__(self, name, inputs, device=None):
        super(SpatialPyramidPoolLayer, self).__init__(
            name, 'spp', 0, inputs=inputs, device=device)
Q
qijun 已提交
1837 1838 1839
        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
1840 1841 1842
            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
1843

Q
qijun 已提交
1844

Z
zhangjinchao01 已提交
1845 1846 1847
@config_layer('batch_norm')
class BatchNormLayer(LayerBase):
    layer_type = 'batch_norm'
Q
qijun 已提交
1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858

    def __init__(self,
                 name,
                 inputs,
                 active_type="linear",
                 bias=True,
                 device=None,
                 use_global_stats=True,
                 moving_average_fraction=0.9,
                 batch_norm_type=None,
                 **xargs):
Z
zhangjinchao01 已提交
1859 1860 1861 1862
        if inputs is None:
            inputs = []
        elif not isinstance(inputs, list):
            inputs = [inputs]
Q
qijun 已提交
1863 1864
        config_assert(
            len(inputs) == 1, "BatchNormLayer must have one and only one input")
Z
zhangjinchao01 已提交
1865 1866 1867 1868 1869 1870 1871 1872
        # 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
1873 1874 1875 1876 1877 1878
            inputs.append(
                Input(
                    inputs[0].input_layer_name,
                    initial_std=0.0,
                    initial_mean=0.0,
                    is_static=True,
1879
                    is_shared=is_shared,
D
dangqingqing 已提交
1880
                    make_layer_name_in_submodel=False, ))
Z
zhangjinchao01 已提交
1881 1882 1883 1884 1885 1886 1887

        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 \
            ((not parallel_nn) or self.config.device > -1) and \
1888
            cudnn_version >= 4007
Z
zhangjinchao01 已提交
1889
        self.layer_type = "cudnn_batch_norm" if use_cudnn else "batch_norm"
Q
qijun 已提交
1890 1891 1892 1893 1894 1895 1896 1897
        super(BatchNormLayer, self).__init__(
            name,
            self.layer_type,
            0,
            active_type=active_type,
            inputs=inputs,
            device=device,
            **xargs)
Z
zhangjinchao01 已提交
1898 1899 1900 1901 1902 1903

        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
1904
        input_layer = self.get_input_layer(0)
Z
zhangjinchao01 已提交
1905
        image_conf = self.config.inputs[0].image_conf
L
Luo Tao 已提交
1906
        parse_image(self.inputs[0].image, input_layer.name, image_conf)
1907 1908 1909 1910 1911

        # Only pass the width and height of input to batch_norm layer 
        # when either of it is non-zero. 
        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
1912
                               image_conf.channels, False)
1913 1914
        else:
            self.set_layer_size(input_layer.size)
Z
zhangjinchao01 已提交
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926

        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
1927

Z
zhangjinchao01 已提交
1928 1929
@config_layer('trans')
class TransLayer(LayerBase):
Q
qijun 已提交
1930 1931 1932 1933 1934 1935
    def __init__(self, name, inputs, device=None):
        super(TransLayer, self).__init__(
            name, 'trans', 0, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 1,
            'TransLayer must have one and only one input')
Z
zhangjinchao01 已提交
1936 1937
        self.set_layer_size(self.get_input_layer(0).size)

Q
qijun 已提交
1938

Z
zhangjinchao01 已提交
1939 1940
@config_layer('resize')
class ResizeLayer(LayerBase):
Q
qijun 已提交
1941 1942 1943 1944 1945 1946 1947
    def __init__(self, name, size, inputs, device=None):
        super(ResizeLayer, self).__init__(
            name, 'resize', size=size, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 1,
            'ResizeLayer must have one and only one input')

Z
zhangjinchao01 已提交
1948 1949 1950

@config_layer('blockexpand')
class BlockExpandLayer(LayerBase):
Q
qijun 已提交
1951 1952 1953
    def __init__(self, name, inputs, device=None):
        super(BlockExpandLayer, self).__init__(
            name, 'blockexpand', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1954 1955
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
Q
qijun 已提交
1956 1957
            parse_block_expand(
                self.inputs[input_index].block_expand, input_layer.name,
Z
zhangjinchao01 已提交
1958
                self.config.inputs[input_index].block_expand_conf)
Q
qijun 已提交
1959 1960 1961 1962 1963 1964
            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
1965

1966 1967
@config_layer('maxout')
class MaxOutLayer(LayerBase):
Q
qijun 已提交
1968 1969 1970
    def __init__(self, name, inputs, **xargs):
        super(MaxOutLayer, self).__init__(
            name, 'maxout', 0, inputs=inputs, **xargs)
1971 1972
        input_layer = self.get_input_layer(0)
        maxout_conf = self.config.inputs[0].maxout_conf
L
Luo Tao 已提交
1973
        parse_maxout(self.inputs[0].maxout, input_layer.name, maxout_conf)
L
Luo Tao 已提交
1974 1975 1976
        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
1977

1978

Z
zhangjinchao01 已提交
1979 1980 1981 1982
# key: cost type
# value: cost class
g_cost_map = {}

Q
qijun 已提交
1983

Z
zhangjinchao01 已提交
1984 1985 1986
# 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
1987 1988
        super(type(cls), cls).__init__(
            name, cost_type, 1, inputs, device=device, coeff=coeff)
Z
zhangjinchao01 已提交
1989

Q
qijun 已提交
1990
    cls = type(class_name, (LayerBase, ), dict(__init__=init))
Z
zhangjinchao01 已提交
1991 1992 1993
    global g_cost_map
    g_cost_map[cost_type] = cls

Q
qijun 已提交
1994

Z
zhangjinchao01 已提交
1995 1996 1997 1998 1999 2000 2001 2002
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
2003
define_cost('SumCost', 'sum_cost')
Z
zhangjinchao01 已提交
2004

Q
qijun 已提交
2005

Z
zhangjinchao01 已提交
2006 2007
@config_layer('hsigmoid')
class HierarchicalSigmoidLayer(LayerBase):
Q
qijun 已提交
2008
    def __init__(self, name, num_classes, inputs, device=None, bias=True):
Z
zhangjinchao01 已提交
2009 2010
        super(HierarchicalSigmoidLayer, self).__init__(
            name, 'hsigmoid', 1, inputs=inputs, device=device)
Q
qijun 已提交
2011 2012 2013
        config_assert(
            len(self.inputs) >= 2,
            'HierarchicalSigmoidLayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2014 2015 2016 2017 2018 2019 2020 2021
        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
2022

Z
zhangjinchao01 已提交
2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046
'''
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
2047 2048


Z
zhangjinchao01 已提交
2049 2050
@config_layer('lambda_cost')
class LambdaCost(LayerBase):
Q
qijun 已提交
2051
    def __init__(self, name, inputs, NDCG_num=5, max_sort_size=-1, device=None):
Z
zhangjinchao01 已提交
2052 2053
        super(LambdaCost, self).__init__(
            name, 'lambda_cost', 1, inputs=inputs, device=device)
Q
qijun 已提交
2054
        config_assert(len(self.inputs) == 2, 'lambdaCost must have 2 inputs')
Z
zhangjinchao01 已提交
2055 2056
        self.config.NDCG_num = NDCG_num
        if max_sort_size != -1:
Q
qijun 已提交
2057 2058 2059
            config_assert(
                NDCG_num <= max_sort_size,
                'NDCG_num must be less than or equal to max_sort_size')
Z
zhangjinchao01 已提交
2060 2061
        self.config.max_sort_size = max_sort_size

Q
qijun 已提交
2062

Z
zhangjinchao01 已提交
2063 2064
@config_layer('nce')
class NCELayer(LayerBase):
Q
qijun 已提交
2065 2066 2067 2068 2069 2070 2071 2072
    def __init__(self,
                 name,
                 num_classes,
                 inputs,
                 num_neg_samples=10,
                 neg_sampling_dist=None,
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2073
        super(NCELayer, self).__init__(name, 'nce', 1, inputs=inputs, **xargs)
Q
qijun 已提交
2074 2075
        config_assert(
            len(self.inputs) >= 2, 'NCELayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2076 2077
        self.config.num_classes = num_classes
        if neg_sampling_dist is not None:
Q
qijun 已提交
2078 2079 2080 2081
            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
2082
            s = sum(neg_sampling_dist)
Q
qijun 已提交
2083 2084 2085
            config_assert(
                abs(s - 1) < 1e-5,
                'The sum of neg_sampling_dist (%s) is not 1' % s)
Z
zhangjinchao01 已提交
2086 2087 2088 2089 2090

            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
2091
        input_layer = self.get_input_layer(num_real_inputs)
Z
zhangjinchao01 已提交
2092 2093 2094 2095
        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
2096 2097
        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
2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110
            # 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
2111
    def __init__(self, name, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
2112 2113
        super(AddToLayer, self).__init__(
            name, 'addto', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2114
        config_assert(len(inputs) > 0, 'inputs cannot be empty for AddToLayer')
Z
zhangjinchao01 已提交
2115 2116 2117 2118 2119
        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
2120

Z
zhangjinchao01 已提交
2121 2122
@config_layer('agent')
class AgentLayer(LayerBase):
Q
qijun 已提交
2123 2124 2125 2126
    def __init__(self, name, size, device=None):
        super(AgentLayer, self).__init__(
            name, 'agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2127 2128 2129

@config_layer('sequence_agent')
class SequenceAgentLayer(LayerBase):
Q
qijun 已提交
2130
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2131 2132 2133
        super(SequenceAgentLayer, self).__init__(
            name, 'sequence_agent', size, inputs=[], device=device)

Q
qijun 已提交
2134

Z
zhangjinchao01 已提交
2135 2136
@config_layer('gather_agent')
class GatherAgentLayer(LayerBase):
Q
qijun 已提交
2137
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2138 2139 2140
        super(GatherAgentLayer, self).__init__(
            name, 'gather_agent', size, inputs=[], device=device)

Q
qijun 已提交
2141

Z
zhangjinchao01 已提交
2142 2143
@config_layer('scatter_agent')
class ScatterAgentLayer(LayerBase):
Q
qijun 已提交
2144
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2145 2146 2147
        super(ScatterAgentLayer, self).__init__(
            name, 'scatter_agent', size, inputs=[], device=device)

Q
qijun 已提交
2148

Z
zhangjinchao01 已提交
2149 2150
@config_layer('sequence_gather_agent')
class SequenceGatherAgentLayer(LayerBase):
Q
qijun 已提交
2151
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2152
        super(SequenceGatherAgentLayer, self).__init__(
Q
qijun 已提交
2153 2154
            name, 'sequence_gather_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2155 2156 2157

@config_layer('sequence_scatter_agent')
class SequenceScatterAgentLayer(LayerBase):
Q
qijun 已提交
2158
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2159
        super(SequenceScatterAgentLayer, self).__init__(
Q
qijun 已提交
2160 2161
            name, 'sequence_scatter_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2162 2163 2164

@config_layer('multiplex')
class MultiplexLayer(LayerBase):
Q
qijun 已提交
2165 2166 2167 2168 2169
    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
2170
        for i in range(1, len(inputs)):
Q
qijun 已提交
2171 2172 2173 2174 2175
            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
2176 2177

@config_func
Q
qijun 已提交
2178 2179 2180
def Link(
        name,
        has_subseq=False, ):
Z
zhangjinchao01 已提交
2181 2182 2183 2184 2185
    link_config = LinkConfig()
    link_config.link_name = name
    link_config.has_subseq = has_subseq
    return link_config

Q
qijun 已提交
2186

Z
zhangjinchao01 已提交
2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200
# memory for recurrent layer group.
# *name* and *size* are actual layer's name and size.
# will return name of the memory,
# 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
Q
qijun 已提交
2201 2202 2203 2204 2205 2206 2207 2208
def Memory(
        name,
        size,
        is_sequence=False,
        boot_layer=None,
        boot_bias=False,
        boot_bias_active_type="",
        boot_with_const_id=None, ):
Z
zhangjinchao01 已提交
2209 2210 2211 2212 2213 2214
    agent_name = name + "+delay1"
    if is_sequence:
        agent_layer = SequenceAgentLayer(agent_name, size)
    else:
        agent_layer = AgentLayer(agent_name, size)
    config_assert(g_current_submodel.is_recurrent_layer_group,
Q
qijun 已提交
2215
                  'Memory should be used in recurrent layer group only')
Z
zhangjinchao01 已提交
2216 2217 2218 2219
    memory = g_current_submodel.memories.add()
    memory.layer_name = MakeLayerNameInSubmodel(name)
    memory.link_name = MakeLayerNameInSubmodel(agent_name)
    memory.is_sequence = is_sequence
Q
qijun 已提交
2220
    options = sum((boot_layer is not None, bool(boot_bias),
Z
zhangjinchao01 已提交
2221
                   boot_with_const_id is not None))
Q
qijun 已提交
2222 2223 2224 2225
    config_assert(
        options <= 1,
        'take one option at most from boot_layer, boot_bias, or boot_with_const_id'
    )
Z
zhangjinchao01 已提交
2226 2227 2228
    if boot_layer is not None:
        boot_layer = MakeLayerNameInParentSubmodel(boot_layer)
        config_assert(boot_layer in g_layer_map,
Q
qijun 已提交
2229 2230
                      'boot_layer "%s" does not correspond to a layer name' %
                      boot_layer)
Z
zhangjinchao01 已提交
2231 2232 2233
        memory.boot_layer_name = boot_layer
    elif boot_bias:
        memory.boot_bias_parameter_name = agent_layer.create_bias_parameter(
Q
qijun 已提交
2234
            boot_bias, size, for_self=False)
Z
zhangjinchao01 已提交
2235 2236 2237 2238 2239
        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
2240

Z
zhangjinchao01 已提交
2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251
# 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
2252 2253 2254 2255
        eos_layer_name="eos_check",
        num_results_per_sample=1,
        beam_size=1,
        log_prob=None, ):
Z
zhangjinchao01 已提交
2256 2257 2258 2259 2260 2261 2262 2263 2264
    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
2265

Z
zhangjinchao01 已提交
2266 2267
@config_layer('expand')
class ExpandLayer(LayerBase):
Q
qijun 已提交
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283
    def __init__(self,
                 name,
                 inputs,
                 trans_type='non-seq',
                 device=None,
                 bias=False):
        super(ExpandLayer, self).__init__(
            name, 'expand', 0, inputs=inputs, device=device)
        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
2284 2285 2286

@config_layer('featmap_expand')
class FeatMapExpandLayer(LayerBase):
Q
qijun 已提交
2287 2288 2289 2290 2291 2292
    def __init__(self, name, inputs, device=None, num_filters=None, bias=False):
        super(FeatMapExpandLayer, self).__init__(
            name, 'featmap_expand', 0, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 1, 'ExpandLayer takes 1 and only 1 inputs')
        if num_filters is not None:
Z
zhangjinchao01 已提交
2293
            self.config.num_filters = num_filters
Q
qijun 已提交
2294
        else:
Z
zhangjinchao01 已提交
2295
            logger.fatal("FeatMapExpandLayer must specify num_filters.")
Q
qijun 已提交
2296
        self.set_layer_size(self.get_input_layer(0).size * num_filters)
Z
zhangjinchao01 已提交
2297 2298 2299 2300


@config_layer('max')
class MaxLayer(LayerBase):
Q
qijun 已提交
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310
    def __init__(self,
                 name,
                 inputs,
                 trans_type='non-seq',
                 active_type='linear',
                 device=None,
                 bias=False,
                 output_max_index=None):
        super(MaxLayer, self).__init__(
            name, 'max', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2311
        config_assert(len(self.inputs) == 1, 'MaxLayer must have 1 input')
Q
qijun 已提交
2312 2313
        self.config.trans_type = trans_type
        self.config.active_type = active_type
Z
zhangjinchao01 已提交
2314 2315 2316 2317
        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)
2318 2319
        if output_max_index is not None:
            self.config.output_max_index = output_max_index
Z
zhangjinchao01 已提交
2320 2321 2322 2323


@config_layer('maxid')
class MaxIdLayer(LayerBase):
Q
qijun 已提交
2324
    def __init__(self, name, inputs, beam_size=None, device=None):
Z
zhangjinchao01 已提交
2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341
        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
2342
    def __init__(self, name, inputs, eos_id, device=None):
Z
zhangjinchao01 已提交
2343 2344 2345
        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
2346
        self.set_layer_size(2)  # boolean output
Z
zhangjinchao01 已提交
2347 2348
        self.config.eos_id = eos_id

Q
qijun 已提交
2349

Z
zhangjinchao01 已提交
2350 2351
@config_layer('seqlastins')
class SequenceLastInstanceLayer(LayerBase):
Q
qijun 已提交
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368
    def __init__(self,
                 name,
                 inputs,
                 active_type='linear',
                 trans_type='non-seq',
                 device=None,
                 bias=False):
        super(SequenceLastInstanceLayer, self).__init__(
            name,
            'seqlastins',
            0,
            inputs=inputs,
            device=device,
            active_type=active_type)
        config_assert(
            len(inputs) == 1, 'SequenceLastInstanceLayer must have 1 input')
        self.config.trans_type = trans_type
Z
zhangjinchao01 已提交
2369 2370 2371 2372 2373
        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
2374

Z
zhangjinchao01 已提交
2375 2376 2377 2378 2379 2380 2381 2382 2383
@config_layer('seqfirstins')
class SequenceFirstInstanceLayer(SequenceLastInstanceLayer):
    def __init__(
            self,
            name,
            inputs,
            active_type='linear',
            trans_type='non-seq',
            device=None,
Q
qijun 已提交
2384 2385 2386 2387 2388 2389 2390 2391
            bias=False, ):
        super(SequenceFirstInstanceLayer, self).__init__(
            name,
            inputs=inputs,
            active_type=active_type,
            device=device,
            bias=bias)
        self.config.trans_type = trans_type
Z
zhangjinchao01 已提交
2392 2393
        self.config.select_first = True

Q
qijun 已提交
2394

Z
zhangjinchao01 已提交
2395 2396
@config_layer('seqconcat')
class SequenceConcatLayer(LayerBase):
Q
qijun 已提交
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
    def __init__(self,
                 name,
                 inputs,
                 active_type='linear',
                 device=None,
                 bias=False):
        super(SequenceConcatLayer, self).__init__(
            name,
            'seqconcat',
            0,
            inputs=inputs,
            device=device,
            active_type=active_type)
        config_assert(
            len(inputs) == 2, 'SequenceConcatLayer must have 2 inputs')
Z
zhangjinchao01 已提交
2412 2413 2414 2415 2416
        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
2417

Z
zhangjinchao01 已提交
2418 2419
@config_layer('seqreshape')
class SequenceReshapeLayer(LayerBase):
Q
qijun 已提交
2420 2421 2422 2423 2424 2425 2426 2427 2428 2429
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_type='linear',
                 device=None,
                 bias=False):
        super(SequenceReshapeLayer, self).__init__(
            name,
            'seqreshape',
Z
zhangjinchao01 已提交
2430
            size,
Q
qijun 已提交
2431 2432 2433 2434 2435
            inputs=inputs,
            device=device,
            active_type=active_type)
        config_assert(
            len(inputs) == 1, 'SequenceReshapeLayer must have 1 inputs')
Z
zhangjinchao01 已提交
2436 2437 2438
        self.set_layer_size(size)
        self.create_bias_parameter(bias, size)

Q
qijun 已提交
2439

Z
zhangjinchao01 已提交
2440 2441
@config_layer('subseq')
class SubSequenceLayer(LayerBase):
Q
qijun 已提交
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
    def __init__(self,
                 name,
                 inputs,
                 active_type='linear',
                 device=None,
                 bias=False):
        super(SubSequenceLayer, self).__init__(
            name,
            'subseq',
            0,
            inputs=inputs,
            device=device,
            active_type=active_type)
Z
zhangjinchao01 已提交
2455 2456 2457 2458 2459 2460
        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
2461

Z
zhangjinchao01 已提交
2462 2463
@config_layer('out_prod')
class OuterProdLayer(LayerBase):
Q
qijun 已提交
2464 2465 2466
    def __init__(self, name, inputs, device=None):
        super(OuterProdLayer, self).__init__(
            name, 'out_prod', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2467 2468 2469 2470 2471
        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
2472

Z
zhangjinchao01 已提交
2473 2474
@config_layer('power')
class PowerLayer(LayerBase):
Q
qijun 已提交
2475 2476 2477
    def __init__(self, name, inputs, device=None):
        super(PowerLayer, self).__init__(
            name, 'power', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2478 2479 2480 2481
        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
2482 2483 2484
        config_assert(1 == input_layer0.size,
                      'The left input is the exponent and should be of size 1')

Z
zhangjinchao01 已提交
2485 2486 2487

@config_layer('slope_intercept')
class SlopeInterceptLayer(LayerBase):
Q
qijun 已提交
2488 2489 2490
    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
2491 2492 2493 2494 2495 2496
        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
2497

Z
zhangjinchao01 已提交
2498 2499
@config_layer('scaling')
class ScalingLayer(LayerBase):
Q
qijun 已提交
2500 2501 2502
    def __init__(self, name, inputs, device=None):
        super(ScalingLayer, self).__init__(
            name, 'scaling', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2503 2504 2505 2506
        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
2507 2508 2509
        config_assert(1 == input_layer0.size,
                      'The left input should be of size 1')

Z
zhangjinchao01 已提交
2510 2511 2512

@config_layer('conv_shift')
class ConvShiftLayer(LayerBase):
Q
qijun 已提交
2513 2514 2515
    def __init__(self, name, inputs, device=None):
        super(ConvShiftLayer, self).__init__(
            name, 'conv_shift', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2516 2517 2518 2519
        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
2520

Z
zhangjinchao01 已提交
2521 2522
@config_layer('convex_comb')
class ConvexCombinationLayer(LayerBase):
Q
qijun 已提交
2523
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
2524
        super(ConvexCombinationLayer, self).__init__(
Q
qijun 已提交
2525 2526 2527
            name, 'convex_comb', size, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 2, 'ConvexCombinationLayer must have 2 inputs')
2528 2529 2530
        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
2531 2532
        self.set_layer_size(size)

Q
qijun 已提交
2533

Z
zhangjinchao01 已提交
2534 2535
@config_layer('interpolation')
class InterpolationLayer(LayerBase):
Q
qijun 已提交
2536
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2537 2538
        super(InterpolationLayer, self).__init__(
            name, 'interpolation', 0, inputs=inputs, device=device)
Q
qijun 已提交
2539 2540
        config_assert(
            len(self.inputs) == 3, 'InterpolationLayer must have 3 inputs')
Z
zhangjinchao01 已提交
2541 2542 2543 2544 2545 2546 2547 2548
        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
2549

L
liaogang 已提交
2550 2551
@config_layer('bilinear_interp')
class BilinearInterpLayer(LayerBase):
Q
qijun 已提交
2552
    def __init__(self, name, inputs, **xargs):
L
liaogang 已提交
2553
        super(BilinearInterpLayer, self).__init__(
L
liaogang 已提交
2554
            name, 'bilinear_interp', 0, inputs=inputs, **xargs)
L
liaogang 已提交
2555
        input_layer = self.get_input_layer(0)
L
Luo Tao 已提交
2556 2557 2558 2559
        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
2560

L
liaogang 已提交
2561

Z
zhangjinchao01 已提交
2562 2563
@config_layer('sum_to_one_norm')
class SumToOneNormLayer(LayerBase):
Q
qijun 已提交
2564
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2565
        super(SumToOneNormLayer, self).__init__(
Q
qijun 已提交
2566 2567 2568
            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
2569 2570 2571
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2572

Z
zhangjinchao01 已提交
2573 2574
@config_layer('cos_vm')
class CosSimVecMatLayer(LayerBase):
Q
qijun 已提交
2575
    def __init__(self, name, size, inputs, cos_scale=1.0, device=None):
Z
zhangjinchao01 已提交
2576
        super(CosSimVecMatLayer, self).__init__(
Q
qijun 已提交
2577
            name, 'cos_vm', size, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2578
        self.config.cos_scale = cos_scale
Q
qijun 已提交
2579 2580
        config_assert(
            len(self.inputs) == 2, 'CosSimVecMatLayer must have 2 inputs')
2581 2582 2583
        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
2584

Q
qijun 已提交
2585

Z
zhangjinchao01 已提交
2586 2587
@config_layer('sampling_id')
class SamplingIdLayer(LayerBase):
Q
qijun 已提交
2588
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2589 2590
        super(SamplingIdLayer, self).__init__(
            name, 'sampling_id', 0, inputs=inputs, device=device)
Q
qijun 已提交
2591 2592
        config_assert(
            len(self.inputs) == 1, 'SamplingIdLayer must have 1 input')
Z
zhangjinchao01 已提交
2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
        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
2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619
    def __init__(self,
                 name,
                 inputs,
                 average_strategy='average',
                 trans_type='non-seq',
                 active_type='linear',
                 device=None,
                 bias=False):
        super(AverageLayer, self).__init__(
            name,
            'average',
            0,
            inputs=inputs,
            device=device,
            active_type=active_type)
Z
zhangjinchao01 已提交
2620
        self.config.average_strategy = average_strategy
Q
qijun 已提交
2621
        self.config.trans_type = trans_type
Z
zhangjinchao01 已提交
2622 2623 2624 2625 2626 2627
        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
2628

Z
zhangjinchao01 已提交
2629 2630
@config_layer('cos')
class CosSimLayer(LayerBase):
Q
qijun 已提交
2631
    def __init__(self, name, inputs, cos_scale=5, device=None):
Z
zhangjinchao01 已提交
2632 2633 2634 2635 2636 2637
        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')
2638
        self.config.cos_scale = cos_scale
Z
zhangjinchao01 已提交
2639 2640 2641 2642


@config_layer('tensor')
class TensorLayer(LayerBase):
Q
qijun 已提交
2643 2644 2645
    def __init__(self, name, size, inputs, device=None, bias=True, **xargs):
        super(TensorLayer, self).__init__(
            name, 'tensor', size, inputs=inputs, device=device, **xargs)
Z
zhangjinchao01 已提交
2646 2647
        config_assert(len(self.inputs) == 2, 'TensorLayer must have 2 inputs')
        config_assert(size > 0, 'size must be positive')
Q
qijun 已提交
2648 2649
        config_assert(inputs[1].parameter_name == None,
                      'second parameter should be None.')
Z
zhangjinchao01 已提交
2650 2651 2652 2653 2654 2655 2656 2657 2658 2659
        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):
Q
qijun 已提交
2660 2661 2662 2663 2664 2665 2666
    def __init__(self,
                 name,
                 inputs,
                 size=0,
                 bias=True,
                 error_clipping_threshold=None,
                 **xargs):
Z
zhangjinchao01 已提交
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
        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)
2684
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2685 2686 2687
                size = operator.calc_output_size(operator_conf.input_sizes)
                if size != 0:
                    self.set_layer_size(size)
2688
            else:
2689 2690
                sz = operator.calc_output_size(operator_conf.input_sizes)
                if sz != 0:
Q
qijun 已提交
2691 2692 2693 2694
                    config_assert(
                        sz == self.config.size,
                        "different inputs have different size: %s vs. %s" %
                        (sz, self.config.size))
Z
zhangjinchao01 已提交
2695 2696 2697 2698
        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
2699 2700 2701
                config_assert(
                    isinstance(input, Projection),
                    "input should be projection or operation")
2702
            if self.config.size == 0 and isinstance(input, Projection):
Z
zhangjinchao01 已提交
2703 2704 2705
                size = input.calc_output_size(input_layer)
                if size != 0:
                    self.set_layer_size(size)
2706
            elif isinstance(input, Projection):
Q
qijun 已提交
2707 2708 2709 2710 2711 2712
                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
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723
        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
2724 2725
                input_config.proj_conf.name = gen_parameter_name(name,
                                                                 input_index)
Z
zhangjinchao01 已提交
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
                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)

2737 2738 2739 2740 2741 2742
        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
2743

2744 2745 2746
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
Z
zhangjinchao01 已提交
2747

2748 2749
        if error_clipping_threshold is not None:
            self.config.error_clipping_threshold = error_clipping_threshold
Z
zhangjinchao01 已提交
2750

Q
qijun 已提交
2751

Z
zhangjinchao01 已提交
2752 2753
# like MixedLayer, but no bias parameter
@config_func
Q
qijun 已提交
2754
def ExpressionLayer(name, inputs, **xargs):
Z
zhangjinchao01 已提交
2755 2756
    MixedLayer(name, inputs, bias=False, **xargs)

Q
qijun 已提交
2757

Z
zhangjinchao01 已提交
2758 2759
@config_layer('concat')
class ConcatenateLayer(LayerBase):
Q
qijun 已提交
2760
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2761
        config_assert(inputs, 'inputs cannot be empty')
2762
        config_assert(not bias, 'ConcatenateLayer cannot support bias.')
Z
zhangjinchao01 已提交
2763 2764 2765 2766 2767 2768
        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
2769
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2770 2771 2772 2773
                size += input_layer.size

        self.set_layer_size(size)

Q
qijun 已提交
2774

Z
zhangjinchao01 已提交
2775 2776 2777
# like concat layer, but each input layer was processed by a Projection.
@config_layer('concat2')
class ConcatenateLayer2(LayerBase):
Q
qijun 已提交
2778
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2779 2780 2781
        config_assert(inputs, 'inputs cannot be empty')
        super(ConcatenateLayer2, self).__init__(
            name, 'concat2', 0, inputs=inputs, **xargs)
2782 2783

        if isinstance(self.inputs[0], ConvProjection):
Q
qijun 已提交
2784 2785 2786 2787 2788 2789
            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.")
2790

Z
zhangjinchao01 已提交
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
        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
2811
                                              input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2812
            dims = input.calc_parameter_dims(input.proj_conf.input_size,
Q
qijun 已提交
2813
                                             input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2814 2815
            self.create_input_parameter(input_index, psize, dims)

2816 2817 2818 2819 2820 2821 2822
        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()

2823 2824 2825
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
2826

Q
qijun 已提交
2827

Z
zhangjinchao01 已提交
2828 2829
@config_layer('recurrent')
class RecurrentLayer(LayerBase):
Q
qijun 已提交
2830
    def __init__(self, name, inputs, reversed=False, bias=True, **xargs):
Y
Yu Yang 已提交
2831 2832
        super(RecurrentLayer, self).__init__(name, 'recurrent', 0, inputs,
                                             **xargs)
Z
zhangjinchao01 已提交
2833 2834 2835 2836 2837 2838 2839 2840 2841
        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
2842

Z
zhangjinchao01 已提交
2843 2844
@config_layer('lstmemory')
class LstmLayer(LayerBase):
Q
qijun 已提交
2845 2846 2847 2848 2849 2850 2851 2852
    def __init__(self,
                 name,
                 inputs,
                 reversed=False,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2853 2854 2855 2856 2857 2858 2859 2860
        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
2861
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2862 2863 2864 2865 2866
        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
2867

Z
zhangjinchao01 已提交
2868 2869
@config_layer('lstm_step')
class LstmStepLayer(LayerBase):
Q
qijun 已提交
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
    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
2880 2881 2882
        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
2883 2884 2885 2886 2887
        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
2888 2889 2890
        self.config.active_state_type = active_state_type
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2891

Z
zhangjinchao01 已提交
2892 2893 2894
# get the specific output from the input layer.
@config_layer('get_output')
class GetOutputLayer(LayerBase):
Q
qijun 已提交
2895 2896 2897 2898
    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
2899 2900 2901 2902
        inputs = self.inputs[0]
        config_assert(inputs.input_layer_argument,
                      'input_layer_argument cannot be empty')

Q
qijun 已提交
2903

Z
zhangjinchao01 已提交
2904 2905
@config_layer('mdlstmemory')
class MDLstmLayer(LayerBase):
Q
qijun 已提交
2906 2907 2908 2909 2910 2911 2912 2913
    def __init__(self,
                 name,
                 inputs,
                 directions=True,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2914 2915
        super(MDLstmLayer, self).__init__(name, 'mdlstmemory', 0, inputs,
                                          **xargs)
Z
zhangjinchao01 已提交
2916 2917 2918 2919
        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
2920 2921
        config_assert(input_layer.size % (3 + dim_num) == 0,
                      "size % (dim_num) should be 0!")
Q
qijun 已提交
2922
        size = input_layer.size / (3 + dim_num)
Z
zhangjinchao01 已提交
2923
        self.set_layer_size(size)
Q
qijun 已提交
2924
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2925 2926 2927
        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
2928 2929
        self.create_input_parameter(0, size * size * (3 + dim_num),
                                    [size, size, 3 + dim_num])
Z
zhangjinchao01 已提交
2930
        #bias includes 3 kinds of peephole, 3+dim_num+2+dim_num
Q
qijun 已提交
2931 2932
        self.create_bias_parameter(bias, size * (5 + 2 * dim_num))

Z
zhangjinchao01 已提交
2933 2934 2935

@config_layer('gated_recurrent')
class GatedRecurrentLayer(LayerBase):
Q
qijun 已提交
2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
    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
2947 2948 2949 2950 2951 2952
        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
2953
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2954 2955 2956
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2957

Z
zhangjinchao01 已提交
2958 2959
@config_layer('gru_step')
class GruStepLayer(LayerBase):
Q
qijun 已提交
2960 2961 2962 2963 2964 2965 2966
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_gate_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2967 2968
        super(GruStepLayer, self).__init__(name, 'gru_step', size, inputs,
                                           **xargs)
Z
zhangjinchao01 已提交
2969 2970 2971
        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
2972 2973 2974 2975 2976
        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
Z
zhangjinchao01 已提交
2977 2978 2979
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2980

Z
zhangjinchao01 已提交
2981 2982 2983 2984 2985 2986 2987
'''
 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
2988 2989


Z
zhangjinchao01 已提交
2990 2991
@config_layer('crf')
class CRFLayer(LayerBase):
Q
qijun 已提交
2992
    def __init__(self, name, size, inputs, coeff=1.0, device=None):
Z
zhangjinchao01 已提交
2993
        super(CRFLayer, self).__init__(name, 'crf', size, inputs, device=device)
Q
qijun 已提交
2994 2995
        config_assert(2 <= len(self.inputs) <= 3,
                      'CRFLayer must have 2 or 3 inputs')
Z
zhangjinchao01 已提交
2996 2997 2998
        self.create_input_parameter(0, size * (size + 2), [size, size + 2])
        self.config.coeff = coeff

Q
qijun 已提交
2999

Z
zhangjinchao01 已提交
3000 3001 3002 3003 3004 3005 3006 3007
'''
 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
3008 3009


Z
zhangjinchao01 已提交
3010 3011
@config_layer('crf_decoding')
class CRFDecodingLayer(LayerBase):
Q
qijun 已提交
3012
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
3013 3014 3015 3016 3017 3018 3019
        super(CRFDecodingLayer, self).__init__(
            name, 'crf_decoding', size, inputs, device=device)
        config_assert(
            len(self.inputs) <= 2,
            'CRFDecodingLayer cannot have more than 2 inputs')
        self.create_input_parameter(0, size * (size + 2), [size, size + 2])

Q
qijun 已提交
3020

Z
zhangjinchao01 已提交
3021 3022
@config_layer('ctc')
class CTCLayer(LayerBase):
Q
qijun 已提交
3023
    def __init__(self, name, size, inputs, norm_by_times=False, device=None):
Z
zhangjinchao01 已提交
3024 3025 3026 3027
        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
3028

3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049
@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
3050 3051
@config_layer('recurrent_layer_group')
class RecurrentLayerGroup(LayerBase):
Q
qijun 已提交
3052
    def __init__(self, name, device=None):
L
Luo Tao 已提交
3053 3054
        global g_pass_height_width
        g_pass_height_width = False
Z
zhangjinchao01 已提交
3055 3056 3057 3058 3059 3060
        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
3061
def Layer(name, type, **xargs):
Z
zhangjinchao01 已提交
3062 3063 3064 3065
    layers = {}
    layers.update(g_cost_map)
    layers.update(g_layer_type_map)
    layer_func = layers.get(type)
Q
qijun 已提交
3066
    config_assert(layer_func, "layer type '%s' not supported." % type)
X
xuwei06 已提交
3067
    return layer_func(name, **xargs)
Z
zhangjinchao01 已提交
3068

Q
qijun 已提交
3069

Z
zhangjinchao01 已提交
3070
@config_func
Q
qijun 已提交
3071
def ParameterHook(type, **kwargs):
Z
zhangjinchao01 已提交
3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083
    if type == 'pruning':
        mask_filename = kwargs.get('mask_filename', None)
        assert mask_filename is not None
        hook = ParameterUpdaterHookConfig()
        hook.type = type
        hook.purning_mask_filename = mask_filename
        return hook
    else:
        return None


@config_func
Q
qijun 已提交
3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105
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,
              update_hooks=None):
Z
zhangjinchao01 已提交
3106 3107 3108 3109 3110 3111 3112

    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
3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123
    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
3124 3125
    config_assert(not momentum or not decay_rate_l1,
                  "momentum and decay_rate_l1 cannot both be non-zero")
3126 3127 3128 3129 3130

    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
3131 3132 3133 3134
    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)
3135

Q
qijun 已提交
3136 3137
    num_batches_regularization = default(num_batches_regularization,
                                         g_default_num_batches_regularization)
3138 3139 3140
    if num_batches_regularization is not None:
        para.num_batches_regularization = int(num_batches_regularization)

Z
zhangjinchao01 已提交
3141 3142 3143 3144 3145 3146
    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
3147 3148
    gradient_clipping_threshold = default(gradient_clipping_threshold,
                                          g_default_gradient_clipping_threshold)
3149 3150
    if gradient_clipping_threshold is not None:
        para.gradient_clipping_threshold = gradient_clipping_threshold
Q
qijun 已提交
3151 3152
    para.initial_strategy = default(initial_strategy,
                                    g_default_initial_strategy)
Z
zhangjinchao01 已提交
3153 3154 3155 3156 3157 3158
    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
3159 3160 3161
            print(
                "Use initial_smart, but dims not set. Initial_smart may not be used in this layer"
            )
Z
zhangjinchao01 已提交
3162 3163 3164 3165
            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)
3166 3167 3168 3169 3170 3171 3172

    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
3173 3174 3175 3176
    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")
3177 3178
    if is_shared is not None:
        para.is_shared = is_shared
Z
zhangjinchao01 已提交
3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199

    update_hooks = default(update_hooks, g_default_update_hooks)

    if update_hooks is not None:
        if hasattr(update_hooks, '__call__'):
            update_hooks = update_hooks(para.name)

        if isinstance(update_hooks, list):
            for hook in update_hooks:
                para.update_hooks.extend([hook])
        else:
            para.update_hooks.extend(update_hooks)

    g_parameter_map[name] = para


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

Q
qijun 已提交
3200

Z
zhangjinchao01 已提交
3201 3202 3203 3204 3205
@config_func
def default_initial_mean(val):
    global g_default_initial_mean
    g_default_initial_mean = val

Q
qijun 已提交
3206

Z
zhangjinchao01 已提交
3207 3208 3209 3210 3211
@config_func
def default_initial_strategy(val):
    global g_default_initial_strategy
    g_default_initial_strategy = val

Q
qijun 已提交
3212

Z
zhangjinchao01 已提交
3213 3214 3215 3216 3217
@config_func
def default_initial_smart(val):
    global g_default_initial_smart
    g_default_initial_smart = val

Q
qijun 已提交
3218

Z
zhangjinchao01 已提交
3219 3220 3221 3222 3223
@config_func
def default_momentum(val):
    global g_default_momentum
    g_default_momentum = val

Q
qijun 已提交
3224

Z
zhangjinchao01 已提交
3225 3226 3227 3228 3229
@config_func
def default_decay_rate(val):
    global g_default_decay_rate
    g_default_decay_rate = val

Q
qijun 已提交
3230

Z
zhangjinchao01 已提交
3231 3232 3233 3234 3235
@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
3236

Z
zhangjinchao01 已提交
3237 3238 3239 3240 3241
@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
3242

Z
zhangjinchao01 已提交
3243 3244 3245 3246 3247
@config_func
def default_device(val):
    global g_default_device
    g_default_device = val

Q
qijun 已提交
3248

Z
zhangjinchao01 已提交
3249 3250 3251 3252 3253
@config_func
def default_update_hooks(val):
    global g_default_update_hooks
    g_default_update_hooks = val

Q
qijun 已提交
3254

Z
zhangjinchao01 已提交
3255 3256 3257 3258 3259
@config_func
def default_compact_func(val):
    global g_default_compact_func
    g_default_compact_func = val

Q
qijun 已提交
3260

Z
zhangjinchao01 已提交
3261 3262 3263 3264 3265
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
3266 3267 3268
        execfile(config_file,
                 make_config_environment(config_file, config_args), local_args)

Z
zhangjinchao01 已提交
3269 3270
    return Import

Q
qijun 已提交
3271

Z
zhangjinchao01 已提交
3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
settings = dict(
    batch_size=None,
    mini_batch_size=None,
    algorithm='async_sgd',
    async_lagged_grad_discard_ratio=1.5,
    learning_method='momentum',
    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
3300 3301 3302
    adam_beta1=0.9,
    adam_beta2=0.999,
    adam_epsilon=1e-8, )
Z
zhangjinchao01 已提交
3303

Q
qijun 已提交
3304
settings_deprecated = dict(usage_ratio=1., )
Z
zhangjinchao01 已提交
3305 3306 3307 3308

trainer_settings = dict(
    save_dir="./output/model",
    init_model_path=None,
Q
qijun 已提交
3309 3310
    start_pass=0, )

Z
zhangjinchao01 已提交
3311 3312 3313 3314 3315

@config_func
def Settings(**args):
    for k, v in args.iteritems():
        if k == "usage_ratio":
Q
qijun 已提交
3316 3317
            logger.warning(
                "Deprecated: define usage_ratio in DataConfig instead")
Z
zhangjinchao01 已提交
3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328
            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
3329

Z
zhangjinchao01 已提交
3330 3331 3332 3333
@config_func
def cluster_config(**args):
    pass

Q
qijun 已提交
3334

Z
zhangjinchao01 已提交
3335 3336 3337 3338 3339 3340 3341 3342 3343
@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
3344

Z
zhangjinchao01 已提交
3345 3346 3347 3348
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
3349

Z
zhangjinchao01 已提交
3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364
        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
3365
        get_config_arg=make_get_config_arg(config_args), )
Z
zhangjinchao01 已提交
3366 3367 3368 3369 3370

    funcs.update(g_extended_config_funcs)

    return funcs

Q
qijun 已提交
3371

Z
zhangjinchao01 已提交
3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387
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
3388

Z
zhangjinchao01 已提交
3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400
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
3401

Z
zhangjinchao01 已提交
3402 3403 3404 3405
def my_fatal(s):
    logger.critical(s)
    raise Exception()

Y
Yu Yang 已提交
3406

3407
_parse_config_hooks = set()
Y
Yu Yang 已提交
3408 3409


3410 3411 3412 3413 3414 3415 3416
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
3417

Y
Yu Yang 已提交
3418

Z
zhangjinchao01 已提交
3419 3420 3421 3422 3423 3424
def parse_config(config_file, config_arg_str):
    '''
    @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
    '''
    init_config_environment()
3425 3426
    for hook in _parse_config_hooks:
        hook()
Z
zhangjinchao01 已提交
3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453

    config_args = {}

    logger.findCaller = find_caller
    logger.fatal = my_fatal

    g_config.model_config.type = "nn"
    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)

    g_config.model_config.type = 'nn'

    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

H
hanchao 已提交
3454 3455 3456 3457 3458 3459 3460 3461 3462 3463
    # for paddle on spark, need support non-file config.
    # you can use parse_config like below:
    #
    # from paddle.trainer.config_parser import parse_config
    # def configs():
    #    #your paddle config code, which is same as config file.
    #
    # config = parse_config(configs, "is_predict=1")
    # # then you get config proto object.
    if hasattr(config_file, '__call__'):
L
Luo Tao 已提交
3464 3465 3466
        config_file.func_globals.update(
            make_config_environment("", config_args))
        config_file()
H
hanchao 已提交
3467
    else:
L
Luo Tao 已提交
3468
        execfile(config_file, make_config_environment(config_file, config_args))
Z
zhangjinchao01 已提交
3469 3470 3471
    for k, v in settings.iteritems():
        if v is None:
            continue
Q
qijun 已提交
3472
        g_config.opt_config.__setattr__(k, v)
Z
zhangjinchao01 已提交
3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498

    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


def parse_config_and_serialize(config_file, config_arg_str):
    try:
        config = parse_config(config_file, config_arg_str)
        #logger.info(config)
        return config.SerializeToString()
    except:
        traceback.print_exc()
        raise

Q
qijun 已提交
3499

Z
zhangjinchao01 已提交
3500 3501 3502 3503 3504 3505 3506 3507
if __name__ == '__main__':
    try:
        config = parse_config(sys.argv[1], '')
        config.SerializeToString()
        __real_print__(str(config))
    except:
        traceback.print_exc()
        raise