config_parser.py 122.7 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,
Q
qijun 已提交
501
            input_layer_argument=None, ):
Z
zhangjinchao01 已提交
502 503 504
        self.add_keys(locals())
        self.input_layer_name = MakeLayerNameInSubmodel(input_layer_name)

Q
qijun 已提交
505

Z
zhangjinchao01 已提交
506 507 508
# Define a projection for iexed layer
@config_class
class Projection(Input):
Q
qijun 已提交
509 510
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
511 512 513
    def __init__(
            self,
            input_layer_name,
Q
qijun 已提交
514
            size=0,  # projection output size
Z
zhangjinchao01 已提交
515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
            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
534
            input_layer_argument=None, ):
Z
zhangjinchao01 已提交
535 536 537 538 539 540 541 542 543 544 545 546 547
        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
548

Z
zhangjinchao01 已提交
549 550
    def calc_parameter_size(self, input_size, output_size):
        raise NotimplementedError
Q
qijun 已提交
551

Z
zhangjinchao01 已提交
552 553 554 555 556 557 558 559 560 561
    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
562

Z
zhangjinchao01 已提交
563 564
    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
565

Z
zhangjinchao01 已提交
566 567 568
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
569

Z
zhangjinchao01 已提交
570 571 572 573 574 575
# 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
576 577 578
    def __init__(self, input_layer_name, offset, **xargs):
        super(IdentityOffsetProjection, self).__init__(input_layer_name,
                                                       **xargs)
Z
zhangjinchao01 已提交
579 580 581 582
        self.proj_conf.offset = offset

    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
583

Z
zhangjinchao01 已提交
584 585 586
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
587

Z
zhangjinchao01 已提交
588 589 590 591 592 593 594
# 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
595

Z
zhangjinchao01 已提交
596 597
    def calc_parameter_size(self, input_size, output_size):
        return output_size
Q
qijun 已提交
598

Z
zhangjinchao01 已提交
599 600 601
    def calc_parameter_dims(self, input_size, output_size):
        return [1, output_size]

L
Luo Tao 已提交
602

X
xuwei06 已提交
603 604 605 606 607 608 609 610 611 612 613 614 615 616
# 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
617

Z
zhangjinchao01 已提交
618 619 620 621 622 623
@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
624

Z
zhangjinchao01 已提交
625 626 627
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
628

Z
zhangjinchao01 已提交
629 630 631 632 633 634
@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
635

Z
zhangjinchao01 已提交
636 637 638
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
639

Z
zhangjinchao01 已提交
640 641 642 643 644 645
@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
646

Z
zhangjinchao01 已提交
647 648 649
    def calc_parameter_dims(self, input_size, output_size):
        return [output_size, input_size]

Q
qijun 已提交
650

Z
zhangjinchao01 已提交
651 652 653 654
@config_class
class ContextProjection(Projection):
    type = 'context'

Q
qijun 已提交
655 656
    def __init__(self, input_layer_name, context_start, context_length,
                 trainable_padding, **xargs):
Z
zhangjinchao01 已提交
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679
        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


680 681 682 683
@config_class
class ConvProjection(Projection):
    type = 'conv'

Q
qijun 已提交
684 685 686 687 688
    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
689 690 691 692 693
        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
694
        parse_conv(conv_conf, input_layer_name, self.proj_conf.conv_conf,
695
                   num_filters)
L
Luo Tao 已提交
696 697 698
        self.proj_conf.output_size = self.proj_conf.conv_conf.output_x * \
                                     self.proj_conf.conv_conf.output_y * \
                                     num_filters
699 700 701 702 703 704 705 706 707

    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
708 709
        gr = self.proj_conf.conv_conf.groups
        return co * ci * fh * fw / gr
710 711 712 713 714 715 716

    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
717

Z
zhangjinchao01 已提交
718 719 720
# Define a operator for mixed layer
@config_class
class Operator(Cfg):
Q
qijun 已提交
721 722
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
723 724
    def __init__(
            self,
Q
qijun 已提交
725
            input_layer_names, ):
Z
zhangjinchao01 已提交
726 727 728 729 730 731 732 733 734 735
        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
736

Z
zhangjinchao01 已提交
737 738 739
@config_class
class DotMulOperator(Operator):
    type = 'dot_mul'
Q
qijun 已提交
740 741 742

    def __init__(self, input_layer_names, scale=None, **xargs):
        super(DotMulOperator, self).__init__(input_layer_names, **xargs)
Z
zhangjinchao01 已提交
743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
        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
761 762 763 764 765 766 767

    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
768 769 770
        if num_filters is not None:
            self.operator_conf.num_filters = num_filters

771 772
        parse_conv(conv_conf,
                   MakeLayerNameInSubmodel(input_layer_names[0]),
Q
qijun 已提交
773
                   self.operator_conf.conv_conf, num_filters)
L
Luo Tao 已提交
774 775 776
        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
777 778 779

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

780 781
    def calc_output_size(self, input_sizes):
        return self.operator_conf.output_size
Z
zhangjinchao01 已提交
782 783 784 785 786


# please refer to the comments in proto/ModelConfig.proto
@config_class
class Conv(Cfg):
Q
qijun 已提交
787 788 789 790 791 792 793 794 795 796 797 798 799
    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
800 801
        self.add_keys(locals())
        if filter_size_y is None:
Q
qijun 已提交
802
            self.filter_size_y = filter_size
Z
zhangjinchao01 已提交
803
        if padding_y is None:
Q
qijun 已提交
804
            self.padding_y = padding
Z
zhangjinchao01 已提交
805
        if stride_y is None:
Q
qijun 已提交
806
            self.stride_y = stride
Z
zhangjinchao01 已提交
807
        if output_x is not None:
Q
qijun 已提交
808 809
            config_assert(output_x <= 0)

Z
zhangjinchao01 已提交
810

L
liaogang 已提交
811 812
@config_class
class BilinearInterp(Cfg):
L
Luo Tao 已提交
813
    def __init__(self, out_size_x=None, out_size_y=None, channels=None):
L
liaogang 已提交
814 815
        self.add_keys(locals())

Q
qijun 已提交
816

Z
zhangjinchao01 已提交
817 818
@config_class
class Pool(Cfg):
D
dangqingqing 已提交
819 820 821 822 823 824 825 826 827 828 829 830
    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
831
        self.add_keys(locals())
Q
qijun 已提交
832 833


Q
qijun 已提交
834
@config_class
Q
qijun 已提交
835
class SpatialPyramidPool(Cfg):
L
Luo Tao 已提交
836
    def __init__(self, pool_type, pyramid_height, channels):
Q
qijun 已提交
837
        self.add_keys(locals())
Z
zhangjinchao01 已提交
838

Q
qijun 已提交
839

Z
zhangjinchao01 已提交
840 841
@config_class
class Norm(Cfg):
Q
qijun 已提交
842 843 844 845 846 847 848 849 850
    def __init__(self,
                 norm_type,
                 channels,
                 size,
                 scale,
                 pow,
                 output_x=None,
                 img_size=None,
                 blocked=None):
Z
zhangjinchao01 已提交
851 852
        self.add_keys(locals())

Q
qijun 已提交
853

Z
zhangjinchao01 已提交
854 855
@config_class
class Image(Cfg):
Q
qijun 已提交
856
    def __init__(self, channels, img_size=None):
Z
zhangjinchao01 已提交
857 858
        self.add_keys(locals())

Q
qijun 已提交
859

Z
zhangjinchao01 已提交
860 861
@config_class
class BlockExpand(Cfg):
Q
qijun 已提交
862 863 864 865 866 867 868 869 870 871 872 873
    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
874 875
        self.add_keys(locals())

Q
qijun 已提交
876

877 878
@config_class
class MaxOut(Cfg):
Q
qijun 已提交
879
    def __init__(self, channels, groups, img_size_x=0, img_size_y=0):
880 881
        self.add_keys(locals())

Q
qijun 已提交
882

Z
zhangjinchao01 已提交
883 884 885 886 887 888 889 890 891 892 893 894 895
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
896 897
    data_config.data_ratio = data_ratio
    data_config.is_main_data = is_main_data
Z
zhangjinchao01 已提交
898

Q
qijun 已提交
899
    usage_ratio = default(usage_ratio, settings_deprecated["usage_ratio"])
Z
zhangjinchao01 已提交
900 901 902 903 904 905
    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
906

Z
zhangjinchao01 已提交
907
@config_func
Q
qijun 已提交
908 909 910 911 912
def SimpleData(files=None,
               feat_dim=None,
               context_len=None,
               buffer_capacity=None,
               **xargs):
Z
zhangjinchao01 已提交
913 914 915 916 917 918 919 920 921 922
    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
923

Z
zhangjinchao01 已提交
924
@config_func
Q
qijun 已提交
925 926 927 928 929 930 931 932 933 934
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
935 936 937
    data_config = DataBase(**xargs)
    data_config.type = 'py'
    if load_data_module in g_py_module_name_list:
Q
qijun 已提交
938

Z
zhangjinchao01 已提交
939 940 941
        def get_path(module):
            m = __import__(load_data_module)
            return os.path.split(os.path.realpath(m.__file__))[0]
Q
qijun 已提交
942

Z
zhangjinchao01 已提交
943 944 945
        # 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
946 947
        module_new_name = "%s_copy_%d" % (load_data_module,
                                          len(g_py_module_name_list))
Z
zhangjinchao01 已提交
948
        g_py_module_name_list.append(module_new_name)
Q
qijun 已提交
949 950 951 952
        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
953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
        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
977

Z
zhangjinchao01 已提交
978
@config_func
Q
qijun 已提交
979 980 981 982 983 984 985
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
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
    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
1006

Z
zhangjinchao01 已提交
1007 1008
#real data for training is actually provided by "sub_data" data providers.
@config_func
Q
qijun 已提交
1009
def MultiData(sub_data=[]):
Z
zhangjinchao01 已提交
1010 1011 1012 1013 1014
    data_config = DataConfig()
    data_config.type = 'multi'
    data_config.sub_data_configs.extend(sub_data)
    return data_config

Q
qijun 已提交
1015

Z
zhangjinchao01 已提交
1016
@config_func
Q
qijun 已提交
1017 1018 1019 1020 1021 1022 1023
def Data(type,
         files=None,
         feat_dim=None,
         slot_dims=None,
         context_len=None,
         buffer_capacity=None,
         **xargs):
Z
zhangjinchao01 已提交
1024 1025 1026 1027 1028 1029 1030 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

    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
1059

L
Luo Tao 已提交
1060 1061
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1062 1063 1064 1065 1066 1067 1068
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
1069

1070
#calcualte image_size based on output_size for de-convolution (ConvTransLayer).
L
Luo Tao 已提交
1071
#It is the reverse function of cnn_output_size
1072
def cnn_image_size(output_size, filter_size, padding, stride, caffe_mode):
L
Luo Tao 已提交
1073 1074 1075
    img_size = (output_size - 1) * stride + filter_size - 2 * padding
    if not caffe_mode:
        img_size = img_size + 1
1076 1077
    return img_size

Q
qijun 已提交
1078

L
Luo Tao 已提交
1079
def get_img_size(input_layer_name, channels):
L
Luo Tao 已提交
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
    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
1098 1099
def parse_pool(pool, input_layer_name, pool_conf):
    pool_conf.pool_type = pool.pool_type
Q
qijun 已提交
1100 1101 1102
    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
1103
                  "['max-projection', 'avg-projection', "
Q
qijun 已提交
1104
                  "'cudnn-max-pool', 'cudnn-avg-pool']" % pool.pool_type)
Z
zhangjinchao01 已提交
1105 1106 1107 1108 1109 1110

    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
1111
    pool_conf.stride_y = default(pool.stride_y, pool_conf.stride)
Z
zhangjinchao01 已提交
1112

L
Luo Tao 已提交
1113
    pool_conf.img_size, pool_conf.img_size_y = \
L
Luo Tao 已提交
1114
        get_img_size(input_layer_name, pool.channels)
Z
zhangjinchao01 已提交
1115

1116
    config_assert(not pool.start, "start is deprecated in pooling.")
Z
zhangjinchao01 已提交
1117

1118
    if pool.padding is not None:
Z
zhangjinchao01 已提交
1119
        pool_conf.padding = pool.padding
1120
    pool_conf.padding_y = default(pool.padding_y, pool_conf.padding)
D
dangqingqing 已提交
1121 1122 1123 1124 1125 1126
    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
1127

Z
zhangjinchao01 已提交
1128

Q
qijun 已提交
1129
def parse_spp(spp, input_layer_name, spp_conf):
L
Luo Tao 已提交
1130
    parse_image(spp, input_layer_name, spp_conf.image_conf)
Q
qijun 已提交
1131 1132
    spp_conf.pool_type = spp.pool_type
    config_assert(spp.pool_type in ['max-projection', 'avg-projection'],
Q
qijun 已提交
1133 1134
                  "pool-type %s is not in "
                  "['max-projection', 'avg-projection']" % spp.pool_type)
Q
qijun 已提交
1135
    spp_conf.pyramid_height = spp.pyramid_height
Q
qijun 已提交
1136

Q
qijun 已提交
1137

Z
zhangjinchao01 已提交
1138 1139
def parse_image(image, input_layer_name, image_conf):
    image_conf.channels = image.channels
L
Luo Tao 已提交
1140
    image_conf.img_size, image_conf.img_size_y = \
L
Luo Tao 已提交
1141
        get_img_size(input_layer_name, image_conf.channels)
Q
qijun 已提交
1142

Z
zhangjinchao01 已提交
1143 1144 1145 1146

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
1147 1148
                  "norm-type %s is not in [rnorm, 'cmrnorm-projection']" %
                  norm.norm_type)
Z
zhangjinchao01 已提交
1149 1150 1151 1152 1153 1154
    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
1155
    norm_conf.img_size, norm_conf.img_size_y = \
L
Luo Tao 已提交
1156
        get_img_size(input_layer_name, norm.channels)
Z
zhangjinchao01 已提交
1157
    norm_conf.output_x = norm_conf.img_size
L
Luo Tao 已提交
1158
    norm_conf.output_y = norm_conf.img_size_y
Z
zhangjinchao01 已提交
1159 1160 1161
    if norm.norm_type in ['cmrnorm-projection']:
        norm_conf.scale /= norm.size
    else:
Q
qijun 已提交
1162 1163
        norm_conf.scale /= norm.size**2

1164

L
Luo Tao 已提交
1165 1166
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1167
def parse_conv(conv, input_layer_name, conv_conf, num_filters, trans=False):
Z
zhangjinchao01 已提交
1168 1169 1170 1171 1172 1173 1174 1175 1176
    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
1177

1178
    if not trans:
1179
        conv_conf.filter_channels = conv.channels / conv.groups
L
Luo Tao 已提交
1180
        conv_conf.img_size, conv_conf.img_size_y = \
L
Luo Tao 已提交
1181
            get_img_size(input_layer_name, conv.channels)
1182
        conv_conf.output_x = cnn_output_size(
Q
qijun 已提交
1183 1184
            conv_conf.img_size, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1185 1186 1187
        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)
1188
    else:
1189
        conv_conf.filter_channels = num_filters / conv.groups
L
Luo Tao 已提交
1190
        conv_conf.output_x, conv_conf.output_y = \
L
Luo Tao 已提交
1191
            get_img_size(input_layer_name, conv.channels)
1192
        conv_conf.img_size = cnn_image_size(
Q
qijun 已提交
1193 1194
            conv_conf.output_x, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1195
        conv_conf.img_size_y = cnn_image_size(
L
Luo Tao 已提交
1196 1197
            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
1198

1199

Z
zhangjinchao01 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
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:
1213
        block_expand_conf.output_x = cnn_output_size(
1214
            block_expand.img_size_x, block_expand.block_x,
1215
            block_expand.padding_x, block_expand.stride_x, False)
Z
zhangjinchao01 已提交
1216 1217

    if block_expand_conf.img_size_y == 0:
1218
        block_expand_conf.output_y = 0
Z
zhangjinchao01 已提交
1219
    else:
1220
        block_expand_conf.output_y = cnn_output_size(
1221
            block_expand.img_size_y, block_expand.block_y,
1222
            block_expand.padding_y, block_expand.stride_y, False)
Z
zhangjinchao01 已提交
1223

Q
qijun 已提交
1224

1225
def parse_maxout(maxout, input_layer_name, maxout_conf):
L
Luo Tao 已提交
1226
    parse_image(maxout, input_layer_name, maxout_conf.image_conf)
1227
    maxout_conf.groups = maxout.groups
1228

Q
qijun 已提交
1229

Z
zhangjinchao01 已提交
1230 1231 1232 1233 1234 1235
# Define an evaluator
@config_func
def Evaluator(
        name,
        type,
        inputs,
Q
qijun 已提交
1236 1237 1238 1239 1240 1241 1242 1243
        chunk_scheme=None,
        num_chunk_types=None,
        classification_threshold=None,
        positive_label=None,
        dict_file=None,
        result_file=None,
        num_results=None,
        delimited=None, ):
Z
zhangjinchao01 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
    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)

1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270
    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
1271

Q
qijun 已提交
1272

Z
zhangjinchao01 已提交
1273 1274 1275 1276 1277
class LayerBase(object):
    def __init__(
            self,
            name,
            type,
Q
qijun 已提交
1278
            size,  # size can be 0. In this case, subclass should set it.
Z
zhangjinchao01 已提交
1279 1280 1281 1282
            inputs,
            device=None,
            active_type="",
            drop_rate=0.,
1283
            coeff=None):
Z
zhangjinchao01 已提交
1284
        config_assert('@' not in name,
Q
qijun 已提交
1285
                      "layer name: %s contain special character @" % name)
Z
zhangjinchao01 已提交
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300
        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()
1301
        assert isinstance(self.config, LayerConfig)
Z
zhangjinchao01 已提交
1302 1303 1304
        self.config.name = name
        self.config.type = type
        self.config.active_type = active_type
1305 1306
        if coeff is not None:
            self.config.coeff = float(coeff)
Z
zhangjinchao01 已提交
1307 1308 1309 1310 1311 1312 1313
        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
1314
        elif g_default_device is not None:
Z
zhangjinchao01 已提交
1315 1316 1317 1318 1319 1320 1321 1322 1323
            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
1324 1325
                    input_layer_name=input,
                    parameter_name=gen_parameter_name(name, input_index))
Z
zhangjinchao01 已提交
1326 1327 1328 1329 1330 1331 1332 1333
                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
1334
                self.operators.append(input)
Z
zhangjinchao01 已提交
1335 1336 1337 1338
                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
1339
                raise ValueError('Wrong type for inputs: %s' % type_of(input))
Z
zhangjinchao01 已提交
1340
            config_assert(input_layer_name in g_layer_map,
Q
qijun 已提交
1341 1342
                          "Unknown input layer '%s' for layer %s" %
                          (input_layer_name, name))
Z
zhangjinchao01 已提交
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353
            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
1354 1355 1356 1357 1358 1359
        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
1360 1361 1362 1363 1364 1365
    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
1366
            bias,  # True/False or BiasCfg
Z
zhangjinchao01 已提交
1367
            size,
Q
qijun 已提交
1368 1369 1370
            dims=None,
            for_self=True,  # whether create bias for layer self
    ):
Z
zhangjinchao01 已提交
1371 1372 1373 1374 1375 1376

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

Q
qijun 已提交
1377 1378 1379
        config_assert(
            type_of(bias) == bool or type_of(bias) == Bias,
            'Incorrect type for bias: %s' % type_of(bias))
Z
zhangjinchao01 已提交
1380 1381 1382 1383 1384 1385 1386 1387 1388

        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:
1389 1390
                assert isinstance(self.config, LayerConfig)

Z
zhangjinchao01 已提交
1391 1392 1393
                Parameter(
                    bias.parameter_name,
                    size,
Q
qijun 已提交
1394 1395
                    self.config.device
                    if self.config.HasField('device') else None,
Z
zhangjinchao01 已提交
1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406
                    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
1407 1408
                    gradient_clipping_threshold=bias.
                    gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1409
                    is_static=bias.is_static,
Q
qijun 已提交
1410
                    is_shared=bias.is_shared, )
Z
zhangjinchao01 已提交
1411 1412 1413 1414 1415
            if for_self:
                self.config.bias_parameter_name = bias.parameter_name
            else:
                return bias.parameter_name

Q
qijun 已提交
1416 1417 1418 1419 1420 1421
    def create_input_parameter(self,
                               input_index,
                               size,
                               dims=None,
                               sparse=None,
                               format=None):
Z
zhangjinchao01 已提交
1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
        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
1436 1437
            config_assert(size == para.size, (
                'Shared parameter "%s" does not ' + 'have same size: %s vs. %s')
Z
zhangjinchao01 已提交
1438 1439
                          % (input_config.parameter_name, para.size, size))

Q
qijun 已提交
1440 1441
            config_assert(dims == para.dims, (
                'Shared parameter "%s" does not ' + 'have same dims: %s vs. %s')
Z
zhangjinchao01 已提交
1442 1443 1444 1445 1446 1447
                          % (input_config.parameter_name, para.dims, dims))
            return

        Parameter(
            input_config.parameter_name,
            size,
1448
            self.config.device if self.config.HasField("device") else None,
Z
zhangjinchao01 已提交
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
            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
1461 1462
            gradient_clipping_threshold=input_config.
            gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1463 1464 1465 1466
            sparse=sparse,
            format=format,
            is_static=input_config.is_static,
            is_shared=input_config.is_shared,
Q
qijun 已提交
1467
            update_hooks=input_config.update_hooks)
Z
zhangjinchao01 已提交
1468 1469 1470 1471 1472 1473 1474 1475 1476

    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
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
    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
1494

Z
zhangjinchao01 已提交
1495 1496
@config_layer('multi_class_cross_entropy_with_selfnorm')
class MultiClassCrossEntropySelfNormCostLayer(LayerBase):
Q
qijun 已提交
1497 1498 1499
    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
1500 1501
        self.config.softmax_selfnorm_alpha = softmax_selfnorm_alpha

Q
qijun 已提交
1502

Z
zhangjinchao01 已提交
1503 1504
@config_layer('fc')
class FCLayer(LayerBase):
Q
qijun 已提交
1505
    def __init__(self, name, size, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
        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
1516 1517
            else:
                sparse = None
Z
zhangjinchao01 已提交
1518

Q
qijun 已提交
1519 1520
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1521 1522
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1523

Z
zhangjinchao01 已提交
1524 1525
@config_layer('selective_fc')
class SelectiveFCLayer(LayerBase):
Q
qijun 已提交
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
    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
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
        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
1556 1557
                          ("if indices of selected columns are not specified, "
                           "selective_fc Layer has at least two inputs"))
Z
zhangjinchao01 已提交
1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569
            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
1570 1571
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1572 1573
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1574

1575 1576
@config_layer('print')
class PrintLayer(LayerBase):
Q
qijun 已提交
1577
    def __init__(self, name, inputs):
1578 1579
        super(PrintLayer, self).__init__(name, 'print', 0, inputs)

Y
yuan 已提交
1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
@config_layer('priorbox')
class PriorBoxLayer(LayerBase):
    def __init__(self, name, inputs, size, min_size, max_size, aspect_ratio, variance):
        super(PriorBoxLayer, self).__init__(name, 'priorbox', 0, inputs)
        config_assert(len(inputs) == 2, 'PriorBoxLayer must have 2 input')
        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
        input_layer0 = self.get_input_layer(0)
        input_layer1 = self.get_input_layer(1)

Q
qijun 已提交
1593

Z
zhangjinchao01 已提交
1594 1595
@config_layer('data')
class DataLayer(LayerBase):
L
Luo Tao 已提交
1596
    def __init__(self, name, size, height=None, width=None, device=None):
Q
qijun 已提交
1597 1598
        super(DataLayer, self).__init__(
            name, 'data', size, inputs=[], device=device)
L
Luo Tao 已提交
1599 1600
        if height and width:
            self.set_layer_height_width(height, width)
Q
qijun 已提交
1601

Z
zhangjinchao01 已提交
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628

'''
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
1629 1630


Z
zhangjinchao01 已提交
1631 1632
@config_layer('data_norm')
class DataNormLayer(LayerBase):
Q
qijun 已提交
1633
    def __init__(self, name, inputs, data_norm_strategy="z-score", device=None):
Z
zhangjinchao01 已提交
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644
        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
1645

Z
zhangjinchao01 已提交
1646 1647 1648
@config_layer('prelu')
class ParameterReluLayer(LayerBase):
    layer_type = 'prelu'
Q
qijun 已提交
1649 1650

    def __init__(self, name, inputs, partial_sum=1, **args):
Z
zhangjinchao01 已提交
1651 1652 1653 1654 1655 1656 1657 1658
        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
1659

Z
zhangjinchao01 已提交
1660 1661 1662
@config_layer('conv')
class ConvLayerBase(LayerBase):
    layer_type = 'conv'
Q
qijun 已提交
1663 1664 1665 1666 1667 1668 1669 1670

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
Z
zhangjinchao01 已提交
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
        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
1687
            (parallel_nn == 0 or self.config.device > -1)):
Z
zhangjinchao01 已提交
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699
            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
1700 1701
            parse_conv(self.inputs[input_index].conv, input_layer.name,
                       conv_conf, num_filters)
Z
zhangjinchao01 已提交
1702 1703
            psize = self.calc_parameter_size(conv_conf)
            self.create_input_parameter(input_index, psize)
L
Luo Tao 已提交
1704 1705
            self.set_cnn_layer(name, conv_conf.output_y, conv_conf.output_x,
                               self.config.num_filters)
Z
zhangjinchao01 已提交
1706 1707 1708 1709 1710 1711 1712 1713 1714 1715

        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
1716

Z
zhangjinchao01 已提交
1717 1718 1719 1720
@config_layer('exconv')
class ConvLayer(ConvLayerBase):
    layer_type = 'exconv'

Q
qijun 已提交
1721

Z
zhangjinchao01 已提交
1722 1723 1724 1725
@config_layer('cudnn_conv')
class ConvLayer(ConvLayerBase):
    layer_type = 'cudnn_conv'

1726 1727 1728 1729

@config_layer('convt')
class ConvTransLayerBase(LayerBase):
    layer_type = 'convt'
Q
qijun 已提交
1730 1731 1732 1733 1734 1735 1736 1737

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
1738
        super(ConvTransLayerBase, self).__init__(
1739 1740 1741 1742 1743 1744 1745 1746
            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))

1747 1748
        # cudnn_convt has not been implemented so use exconvt only
        self.layer_type = "exconvt"
1749 1750 1751 1752 1753 1754 1755 1756
        # 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)
1757
            parse_conv(
1758 1759
                self.inputs[input_index].conv,
                input_layer.name,
1760
                self.config.inputs[input_index].conv_conf,
1761
                num_filters,
1762
                trans=True)
1763 1764 1765 1766 1767
            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
1768
                (conv_conf.img_size**2) * self.config.num_filters)
1769 1770 1771 1772 1773 1774 1775

        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):
1776
        return conv_conf.channels * conv_conf.filter_channels \
1777 1778
                    * (conv_conf.filter_size * conv_conf.filter_size_y)

Q
qijun 已提交
1779

1780 1781 1782 1783
@config_layer('exconvt')
class ConvTransLayer(ConvTransLayerBase):
    layer_type = 'exconvt'

Q
qijun 已提交
1784

Z
zhangjinchao01 已提交
1785 1786
@config_layer('norm')
class NormLayer(LayerBase):
Q
qijun 已提交
1787 1788 1789
    def __init__(self, name, inputs, device=None):
        super(NormLayer, self).__init__(
            name, 'norm', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1790 1791 1792
        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
1793 1794 1795 1796
            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
1797

Z
zhangjinchao01 已提交
1798 1799 1800

@config_layer('pool')
class PoolLayer(LayerBase):
Q
qijun 已提交
1801 1802 1803
    def __init__(self, name, inputs, device=None):
        super(PoolLayer, self).__init__(
            name, 'pool', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1804 1805 1806
        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
1807 1808 1809 1810
            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
1811

Z
zhangjinchao01 已提交
1812

Q
qijun 已提交
1813 1814
@config_layer('spp')
class SpatialPyramidPoolLayer(LayerBase):
Q
qijun 已提交
1815 1816 1817
    def __init__(self, name, inputs, device=None):
        super(SpatialPyramidPoolLayer, self).__init__(
            name, 'spp', 0, inputs=inputs, device=device)
Q
qijun 已提交
1818 1819 1820
        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
1821 1822 1823
            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
1824

Q
qijun 已提交
1825

Z
zhangjinchao01 已提交
1826 1827 1828
@config_layer('batch_norm')
class BatchNormLayer(LayerBase):
    layer_type = 'batch_norm'
Q
qijun 已提交
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839

    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
1840 1841 1842 1843
        if inputs is None:
            inputs = []
        elif not isinstance(inputs, list):
            inputs = [inputs]
Q
qijun 已提交
1844 1845
        config_assert(
            len(inputs) == 1, "BatchNormLayer must have one and only one input")
Z
zhangjinchao01 已提交
1846 1847 1848 1849 1850 1851 1852 1853
        # 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
1854 1855 1856 1857 1858 1859 1860
            inputs.append(
                Input(
                    inputs[0].input_layer_name,
                    initial_std=0.0,
                    initial_mean=0.0,
                    is_static=True,
                    is_shared=is_shared, ))
Z
zhangjinchao01 已提交
1861 1862 1863 1864 1865 1866 1867

        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 \
1868
            cudnn_version >= 4007
Z
zhangjinchao01 已提交
1869
        self.layer_type = "cudnn_batch_norm" if use_cudnn else "batch_norm"
Q
qijun 已提交
1870 1871 1872 1873 1874 1875 1876 1877
        super(BatchNormLayer, self).__init__(
            name,
            self.layer_type,
            0,
            active_type=active_type,
            inputs=inputs,
            device=device,
            **xargs)
Z
zhangjinchao01 已提交
1878 1879 1880 1881 1882 1883

        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
1884
        input_layer = self.get_input_layer(0)
Z
zhangjinchao01 已提交
1885
        image_conf = self.config.inputs[0].image_conf
L
Luo Tao 已提交
1886 1887
        parse_image(self.inputs[0].image, input_layer.name, image_conf)
        self.set_cnn_layer(name, image_conf.img_size_y, image_conf.img_size,
L
Luo Tao 已提交
1888
                           image_conf.channels, False)
Z
zhangjinchao01 已提交
1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900

        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
1901

Z
zhangjinchao01 已提交
1902 1903
@config_layer('trans')
class TransLayer(LayerBase):
Q
qijun 已提交
1904 1905 1906 1907 1908 1909
    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
1910 1911
        self.set_layer_size(self.get_input_layer(0).size)

Q
qijun 已提交
1912

Z
zhangjinchao01 已提交
1913 1914
@config_layer('resize')
class ResizeLayer(LayerBase):
Q
qijun 已提交
1915 1916 1917 1918 1919 1920 1921
    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
1922 1923 1924

@config_layer('blockexpand')
class BlockExpandLayer(LayerBase):
Q
qijun 已提交
1925 1926 1927
    def __init__(self, name, inputs, device=None):
        super(BlockExpandLayer, self).__init__(
            name, 'blockexpand', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1928 1929
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
Q
qijun 已提交
1930 1931
            parse_block_expand(
                self.inputs[input_index].block_expand, input_layer.name,
Z
zhangjinchao01 已提交
1932
                self.config.inputs[input_index].block_expand_conf)
Q
qijun 已提交
1933 1934 1935 1936 1937 1938
            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
1939

1940 1941
@config_layer('maxout')
class MaxOutLayer(LayerBase):
Q
qijun 已提交
1942 1943 1944
    def __init__(self, name, inputs, **xargs):
        super(MaxOutLayer, self).__init__(
            name, 'maxout', 0, inputs=inputs, **xargs)
1945 1946
        input_layer = self.get_input_layer(0)
        maxout_conf = self.config.inputs[0].maxout_conf
L
Luo Tao 已提交
1947
        parse_maxout(self.inputs[0].maxout, input_layer.name, maxout_conf)
L
Luo Tao 已提交
1948 1949 1950
        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
1951

1952

Z
zhangjinchao01 已提交
1953 1954 1955 1956
# key: cost type
# value: cost class
g_cost_map = {}

Q
qijun 已提交
1957

Z
zhangjinchao01 已提交
1958 1959 1960
# 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
1961 1962
        super(type(cls), cls).__init__(
            name, cost_type, 1, inputs, device=device, coeff=coeff)
Z
zhangjinchao01 已提交
1963

Q
qijun 已提交
1964
    cls = type(class_name, (LayerBase, ), dict(__init__=init))
Z
zhangjinchao01 已提交
1965 1966 1967
    global g_cost_map
    g_cost_map[cost_type] = cls

Q
qijun 已提交
1968

Z
zhangjinchao01 已提交
1969 1970 1971 1972 1973 1974 1975 1976
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
1977
define_cost('SumCost', 'sum_cost')
Z
zhangjinchao01 已提交
1978

Q
qijun 已提交
1979

Z
zhangjinchao01 已提交
1980 1981
@config_layer('hsigmoid')
class HierarchicalSigmoidLayer(LayerBase):
Q
qijun 已提交
1982
    def __init__(self, name, num_classes, inputs, device=None, bias=True):
Z
zhangjinchao01 已提交
1983 1984
        super(HierarchicalSigmoidLayer, self).__init__(
            name, 'hsigmoid', 1, inputs=inputs, device=device)
Q
qijun 已提交
1985 1986 1987
        config_assert(
            len(self.inputs) >= 2,
            'HierarchicalSigmoidLayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
1988 1989 1990 1991 1992 1993 1994 1995
        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
1996

Z
zhangjinchao01 已提交
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
'''
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
2021 2022


Z
zhangjinchao01 已提交
2023 2024
@config_layer('lambda_cost')
class LambdaCost(LayerBase):
Q
qijun 已提交
2025
    def __init__(self, name, inputs, NDCG_num=5, max_sort_size=-1, device=None):
Z
zhangjinchao01 已提交
2026 2027
        super(LambdaCost, self).__init__(
            name, 'lambda_cost', 1, inputs=inputs, device=device)
Q
qijun 已提交
2028
        config_assert(len(self.inputs) == 2, 'lambdaCost must have 2 inputs')
Z
zhangjinchao01 已提交
2029 2030
        self.config.NDCG_num = NDCG_num
        if max_sort_size != -1:
Q
qijun 已提交
2031 2032 2033
            config_assert(
                NDCG_num <= max_sort_size,
                'NDCG_num must be less than or equal to max_sort_size')
Z
zhangjinchao01 已提交
2034 2035
        self.config.max_sort_size = max_sort_size

Q
qijun 已提交
2036

Z
zhangjinchao01 已提交
2037 2038
@config_layer('nce')
class NCELayer(LayerBase):
Q
qijun 已提交
2039 2040 2041 2042 2043 2044 2045 2046
    def __init__(self,
                 name,
                 num_classes,
                 inputs,
                 num_neg_samples=10,
                 neg_sampling_dist=None,
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2047
        super(NCELayer, self).__init__(name, 'nce', 1, inputs=inputs, **xargs)
Q
qijun 已提交
2048 2049
        config_assert(
            len(self.inputs) >= 2, 'NCELayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2050 2051
        self.config.num_classes = num_classes
        if neg_sampling_dist is not None:
Q
qijun 已提交
2052 2053 2054 2055
            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
2056
            s = sum(neg_sampling_dist)
Q
qijun 已提交
2057 2058 2059
            config_assert(
                abs(s - 1) < 1e-5,
                'The sum of neg_sampling_dist (%s) is not 1' % s)
Z
zhangjinchao01 已提交
2060 2061 2062 2063 2064

            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
2065
        input_layer = self.get_input_layer(num_real_inputs)
Z
zhangjinchao01 已提交
2066 2067 2068 2069
        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
2070 2071
        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
2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
            # 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
2085
    def __init__(self, name, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
2086 2087
        super(AddToLayer, self).__init__(
            name, 'addto', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2088
        config_assert(len(inputs) > 0, 'inputs cannot be empty for AddToLayer')
Z
zhangjinchao01 已提交
2089 2090 2091 2092 2093
        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
2094

Z
zhangjinchao01 已提交
2095 2096
@config_layer('agent')
class AgentLayer(LayerBase):
Q
qijun 已提交
2097 2098 2099 2100
    def __init__(self, name, size, device=None):
        super(AgentLayer, self).__init__(
            name, 'agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2101 2102 2103

@config_layer('sequence_agent')
class SequenceAgentLayer(LayerBase):
Q
qijun 已提交
2104
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2105 2106 2107
        super(SequenceAgentLayer, self).__init__(
            name, 'sequence_agent', size, inputs=[], device=device)

Q
qijun 已提交
2108

Z
zhangjinchao01 已提交
2109 2110
@config_layer('gather_agent')
class GatherAgentLayer(LayerBase):
Q
qijun 已提交
2111
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2112 2113 2114
        super(GatherAgentLayer, self).__init__(
            name, 'gather_agent', size, inputs=[], device=device)

Q
qijun 已提交
2115

Z
zhangjinchao01 已提交
2116 2117
@config_layer('scatter_agent')
class ScatterAgentLayer(LayerBase):
Q
qijun 已提交
2118
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2119 2120 2121
        super(ScatterAgentLayer, self).__init__(
            name, 'scatter_agent', size, inputs=[], device=device)

Q
qijun 已提交
2122

Z
zhangjinchao01 已提交
2123 2124
@config_layer('sequence_gather_agent')
class SequenceGatherAgentLayer(LayerBase):
Q
qijun 已提交
2125
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2126
        super(SequenceGatherAgentLayer, self).__init__(
Q
qijun 已提交
2127 2128
            name, 'sequence_gather_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2129 2130 2131

@config_layer('sequence_scatter_agent')
class SequenceScatterAgentLayer(LayerBase):
Q
qijun 已提交
2132
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2133
        super(SequenceScatterAgentLayer, self).__init__(
Q
qijun 已提交
2134 2135
            name, 'sequence_scatter_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2136 2137 2138

@config_layer('multiplex')
class MultiplexLayer(LayerBase):
Q
qijun 已提交
2139 2140 2141 2142 2143
    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
2144
        for i in range(1, len(inputs)):
Q
qijun 已提交
2145 2146 2147 2148 2149
            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
2150 2151

@config_func
Q
qijun 已提交
2152 2153 2154
def Link(
        name,
        has_subseq=False, ):
Z
zhangjinchao01 已提交
2155 2156 2157 2158 2159
    link_config = LinkConfig()
    link_config.link_name = name
    link_config.has_subseq = has_subseq
    return link_config

Q
qijun 已提交
2160

Z
zhangjinchao01 已提交
2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174
# 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
2175 2176 2177 2178 2179 2180 2181 2182
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
2183 2184 2185 2186 2187 2188
    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
2189
                  'Memory should be used in recurrent layer group only')
Z
zhangjinchao01 已提交
2190 2191 2192 2193
    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
2194
    options = sum((boot_layer is not None, bool(boot_bias),
Z
zhangjinchao01 已提交
2195
                   boot_with_const_id is not None))
Q
qijun 已提交
2196 2197 2198 2199
    config_assert(
        options <= 1,
        'take one option at most from boot_layer, boot_bias, or boot_with_const_id'
    )
Z
zhangjinchao01 已提交
2200 2201 2202
    if boot_layer is not None:
        boot_layer = MakeLayerNameInParentSubmodel(boot_layer)
        config_assert(boot_layer in g_layer_map,
Q
qijun 已提交
2203 2204
                      'boot_layer "%s" does not correspond to a layer name' %
                      boot_layer)
Z
zhangjinchao01 已提交
2205 2206 2207
        memory.boot_layer_name = boot_layer
    elif boot_bias:
        memory.boot_bias_parameter_name = agent_layer.create_bias_parameter(
Q
qijun 已提交
2208
            boot_bias, size, for_self=False)
Z
zhangjinchao01 已提交
2209 2210 2211 2212 2213
        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
2214

Z
zhangjinchao01 已提交
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225
# 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
2226 2227 2228 2229
        eos_layer_name="eos_check",
        num_results_per_sample=1,
        beam_size=1,
        log_prob=None, ):
Z
zhangjinchao01 已提交
2230 2231 2232 2233 2234 2235 2236 2237 2238
    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
2239

Z
zhangjinchao01 已提交
2240 2241
@config_layer('expand')
class ExpandLayer(LayerBase):
Q
qijun 已提交
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257
    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
2258 2259 2260

@config_layer('featmap_expand')
class FeatMapExpandLayer(LayerBase):
Q
qijun 已提交
2261 2262 2263 2264 2265 2266
    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
2267
            self.config.num_filters = num_filters
Q
qijun 已提交
2268
        else:
Z
zhangjinchao01 已提交
2269
            logger.fatal("FeatMapExpandLayer must specify num_filters.")
Q
qijun 已提交
2270
        self.set_layer_size(self.get_input_layer(0).size * num_filters)
Z
zhangjinchao01 已提交
2271 2272 2273 2274


@config_layer('max')
class MaxLayer(LayerBase):
Q
qijun 已提交
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
    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
2285
        config_assert(len(self.inputs) == 1, 'MaxLayer must have 1 input')
Q
qijun 已提交
2286 2287
        self.config.trans_type = trans_type
        self.config.active_type = active_type
Z
zhangjinchao01 已提交
2288 2289 2290 2291
        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)
2292 2293
        if output_max_index is not None:
            self.config.output_max_index = output_max_index
Z
zhangjinchao01 已提交
2294 2295 2296 2297


@config_layer('maxid')
class MaxIdLayer(LayerBase):
Q
qijun 已提交
2298
    def __init__(self, name, inputs, beam_size=None, device=None):
Z
zhangjinchao01 已提交
2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315
        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
2316
    def __init__(self, name, inputs, eos_id, device=None):
Z
zhangjinchao01 已提交
2317 2318 2319
        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
2320
        self.set_layer_size(2)  # boolean output
Z
zhangjinchao01 已提交
2321 2322
        self.config.eos_id = eos_id

Q
qijun 已提交
2323

Z
zhangjinchao01 已提交
2324 2325
@config_layer('seqlastins')
class SequenceLastInstanceLayer(LayerBase):
Q
qijun 已提交
2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342
    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
2343 2344 2345 2346 2347
        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
2348

Z
zhangjinchao01 已提交
2349 2350 2351 2352 2353 2354 2355 2356 2357
@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
2358 2359 2360 2361 2362 2363 2364 2365
            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
2366 2367
        self.config.select_first = True

Q
qijun 已提交
2368

Z
zhangjinchao01 已提交
2369 2370
@config_layer('seqconcat')
class SequenceConcatLayer(LayerBase):
Q
qijun 已提交
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385
    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
2386 2387 2388 2389 2390
        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
2391

Z
zhangjinchao01 已提交
2392 2393
@config_layer('seqreshape')
class SequenceReshapeLayer(LayerBase):
Q
qijun 已提交
2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_type='linear',
                 device=None,
                 bias=False):
        super(SequenceReshapeLayer, self).__init__(
            name,
            'seqreshape',
Z
zhangjinchao01 已提交
2404
            size,
Q
qijun 已提交
2405 2406 2407 2408 2409
            inputs=inputs,
            device=device,
            active_type=active_type)
        config_assert(
            len(inputs) == 1, 'SequenceReshapeLayer must have 1 inputs')
Z
zhangjinchao01 已提交
2410 2411 2412
        self.set_layer_size(size)
        self.create_bias_parameter(bias, size)

Q
qijun 已提交
2413

Z
zhangjinchao01 已提交
2414 2415
@config_layer('subseq')
class SubSequenceLayer(LayerBase):
Q
qijun 已提交
2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428
    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
2429 2430 2431 2432 2433 2434
        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
2435

Z
zhangjinchao01 已提交
2436 2437
@config_layer('out_prod')
class OuterProdLayer(LayerBase):
Q
qijun 已提交
2438 2439 2440
    def __init__(self, name, inputs, device=None):
        super(OuterProdLayer, self).__init__(
            name, 'out_prod', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2441 2442 2443 2444 2445
        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
2446

Z
zhangjinchao01 已提交
2447 2448
@config_layer('power')
class PowerLayer(LayerBase):
Q
qijun 已提交
2449 2450 2451
    def __init__(self, name, inputs, device=None):
        super(PowerLayer, self).__init__(
            name, 'power', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2452 2453 2454 2455
        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
2456 2457 2458
        config_assert(1 == input_layer0.size,
                      'The left input is the exponent and should be of size 1')

Z
zhangjinchao01 已提交
2459 2460 2461

@config_layer('slope_intercept')
class SlopeInterceptLayer(LayerBase):
Q
qijun 已提交
2462 2463 2464
    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
2465 2466 2467 2468 2469 2470
        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
2471

Z
zhangjinchao01 已提交
2472 2473
@config_layer('scaling')
class ScalingLayer(LayerBase):
Q
qijun 已提交
2474 2475 2476
    def __init__(self, name, inputs, device=None):
        super(ScalingLayer, self).__init__(
            name, 'scaling', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2477 2478 2479 2480
        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
2481 2482 2483
        config_assert(1 == input_layer0.size,
                      'The left input should be of size 1')

Z
zhangjinchao01 已提交
2484 2485 2486

@config_layer('conv_shift')
class ConvShiftLayer(LayerBase):
Q
qijun 已提交
2487 2488 2489
    def __init__(self, name, inputs, device=None):
        super(ConvShiftLayer, self).__init__(
            name, 'conv_shift', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2490 2491 2492 2493
        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
2494

Z
zhangjinchao01 已提交
2495 2496
@config_layer('convex_comb')
class ConvexCombinationLayer(LayerBase):
Q
qijun 已提交
2497
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
2498
        super(ConvexCombinationLayer, self).__init__(
Q
qijun 已提交
2499 2500 2501
            name, 'convex_comb', size, inputs=inputs, device=device)
        config_assert(
            len(self.inputs) == 2, 'ConvexCombinationLayer must have 2 inputs')
2502 2503 2504
        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
2505 2506
        self.set_layer_size(size)

Q
qijun 已提交
2507

Z
zhangjinchao01 已提交
2508 2509
@config_layer('interpolation')
class InterpolationLayer(LayerBase):
Q
qijun 已提交
2510
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2511 2512
        super(InterpolationLayer, self).__init__(
            name, 'interpolation', 0, inputs=inputs, device=device)
Q
qijun 已提交
2513 2514
        config_assert(
            len(self.inputs) == 3, 'InterpolationLayer must have 3 inputs')
Z
zhangjinchao01 已提交
2515 2516 2517 2518 2519 2520 2521 2522
        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
2523

L
liaogang 已提交
2524 2525
@config_layer('bilinear_interp')
class BilinearInterpLayer(LayerBase):
Q
qijun 已提交
2526
    def __init__(self, name, inputs, **xargs):
L
liaogang 已提交
2527
        super(BilinearInterpLayer, self).__init__(
L
liaogang 已提交
2528
            name, 'bilinear_interp', 0, inputs=inputs, **xargs)
L
liaogang 已提交
2529
        input_layer = self.get_input_layer(0)
L
Luo Tao 已提交
2530 2531 2532 2533
        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
2534

L
liaogang 已提交
2535

Z
zhangjinchao01 已提交
2536 2537
@config_layer('sum_to_one_norm')
class SumToOneNormLayer(LayerBase):
Q
qijun 已提交
2538
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2539
        super(SumToOneNormLayer, self).__init__(
Q
qijun 已提交
2540 2541 2542
            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
2543 2544 2545
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2546

Z
zhangjinchao01 已提交
2547 2548
@config_layer('cos_vm')
class CosSimVecMatLayer(LayerBase):
Q
qijun 已提交
2549
    def __init__(self, name, size, inputs, cos_scale=1.0, device=None):
Z
zhangjinchao01 已提交
2550
        super(CosSimVecMatLayer, self).__init__(
Q
qijun 已提交
2551
            name, 'cos_vm', size, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2552
        self.config.cos_scale = cos_scale
Q
qijun 已提交
2553 2554
        config_assert(
            len(self.inputs) == 2, 'CosSimVecMatLayer must have 2 inputs')
2555 2556 2557
        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
2558

Q
qijun 已提交
2559

Z
zhangjinchao01 已提交
2560 2561
@config_layer('sampling_id')
class SamplingIdLayer(LayerBase):
Q
qijun 已提交
2562
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2563 2564
        super(SamplingIdLayer, self).__init__(
            name, 'sampling_id', 0, inputs=inputs, device=device)
Q
qijun 已提交
2565 2566
        config_assert(
            len(self.inputs) == 1, 'SamplingIdLayer must have 1 input')
Z
zhangjinchao01 已提交
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578
        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
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593
    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
2594
        self.config.average_strategy = average_strategy
Q
qijun 已提交
2595
        self.config.trans_type = trans_type
Z
zhangjinchao01 已提交
2596 2597 2598 2599 2600 2601
        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
2602

Z
zhangjinchao01 已提交
2603 2604
@config_layer('cos')
class CosSimLayer(LayerBase):
Q
qijun 已提交
2605
    def __init__(self, name, inputs, cos_scale=5, device=None):
Z
zhangjinchao01 已提交
2606 2607 2608 2609 2610 2611
        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')
2612
        self.config.cos_scale = cos_scale
Z
zhangjinchao01 已提交
2613 2614 2615 2616


@config_layer('tensor')
class TensorLayer(LayerBase):
Q
qijun 已提交
2617 2618 2619
    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
2620 2621
        config_assert(len(self.inputs) == 2, 'TensorLayer must have 2 inputs')
        config_assert(size > 0, 'size must be positive')
Q
qijun 已提交
2622 2623
        config_assert(inputs[1].parameter_name == None,
                      'second parameter should be None.')
Z
zhangjinchao01 已提交
2624 2625 2626 2627 2628 2629 2630 2631 2632 2633
        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
2634 2635 2636 2637 2638 2639 2640
    def __init__(self,
                 name,
                 inputs,
                 size=0,
                 bias=True,
                 error_clipping_threshold=None,
                 **xargs):
Z
zhangjinchao01 已提交
2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657
        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)
2658
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2659 2660 2661
                size = operator.calc_output_size(operator_conf.input_sizes)
                if size != 0:
                    self.set_layer_size(size)
2662
            else:
2663 2664
                sz = operator.calc_output_size(operator_conf.input_sizes)
                if sz != 0:
Q
qijun 已提交
2665 2666 2667 2668
                    config_assert(
                        sz == self.config.size,
                        "different inputs have different size: %s vs. %s" %
                        (sz, self.config.size))
Z
zhangjinchao01 已提交
2669 2670 2671 2672
        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
2673 2674 2675
                config_assert(
                    isinstance(input, Projection),
                    "input should be projection or operation")
2676
            if self.config.size == 0 and isinstance(input, Projection):
Z
zhangjinchao01 已提交
2677 2678 2679
                size = input.calc_output_size(input_layer)
                if size != 0:
                    self.set_layer_size(size)
2680
            elif isinstance(input, Projection):
Q
qijun 已提交
2681 2682 2683 2684 2685 2686
                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
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697
        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
2698 2699
                input_config.proj_conf.name = gen_parameter_name(name,
                                                                 input_index)
Z
zhangjinchao01 已提交
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710
                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)

2711 2712 2713 2714 2715 2716
        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
2717

2718 2719 2720
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
Z
zhangjinchao01 已提交
2721

2722 2723
        if error_clipping_threshold is not None:
            self.config.error_clipping_threshold = error_clipping_threshold
Z
zhangjinchao01 已提交
2724

Q
qijun 已提交
2725

Z
zhangjinchao01 已提交
2726 2727
# like MixedLayer, but no bias parameter
@config_func
Q
qijun 已提交
2728
def ExpressionLayer(name, inputs, **xargs):
Z
zhangjinchao01 已提交
2729 2730
    MixedLayer(name, inputs, bias=False, **xargs)

Q
qijun 已提交
2731

Z
zhangjinchao01 已提交
2732 2733
@config_layer('concat')
class ConcatenateLayer(LayerBase):
Q
qijun 已提交
2734
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2735
        config_assert(inputs, 'inputs cannot be empty')
2736
        config_assert(not bias, 'ConcatenateLayer cannot support bias.')
Z
zhangjinchao01 已提交
2737 2738 2739 2740 2741 2742
        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
2743
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2744 2745 2746 2747
                size += input_layer.size

        self.set_layer_size(size)

Q
qijun 已提交
2748

Z
zhangjinchao01 已提交
2749 2750 2751
# like concat layer, but each input layer was processed by a Projection.
@config_layer('concat2')
class ConcatenateLayer2(LayerBase):
Q
qijun 已提交
2752
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2753 2754 2755
        config_assert(inputs, 'inputs cannot be empty')
        super(ConcatenateLayer2, self).__init__(
            name, 'concat2', 0, inputs=inputs, **xargs)
2756 2757

        if isinstance(self.inputs[0], ConvProjection):
Q
qijun 已提交
2758 2759 2760 2761 2762 2763
            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.")
2764

Z
zhangjinchao01 已提交
2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784
        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
2785
                                              input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2786
            dims = input.calc_parameter_dims(input.proj_conf.input_size,
Q
qijun 已提交
2787
                                             input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2788 2789
            self.create_input_parameter(input_index, psize, dims)

2790 2791 2792 2793 2794 2795 2796
        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()

2797 2798 2799
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
2800

Q
qijun 已提交
2801

Z
zhangjinchao01 已提交
2802 2803
@config_layer('recurrent')
class RecurrentLayer(LayerBase):
Q
qijun 已提交
2804
    def __init__(self, name, inputs, reversed=False, bias=True, **xargs):
Y
Yu Yang 已提交
2805 2806
        super(RecurrentLayer, self).__init__(name, 'recurrent', 0, inputs,
                                             **xargs)
Z
zhangjinchao01 已提交
2807 2808 2809 2810 2811 2812 2813 2814 2815
        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
2816

Z
zhangjinchao01 已提交
2817 2818
@config_layer('lstmemory')
class LstmLayer(LayerBase):
Q
qijun 已提交
2819 2820 2821 2822 2823 2824 2825 2826
    def __init__(self,
                 name,
                 inputs,
                 reversed=False,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2827 2828 2829 2830 2831 2832 2833 2834
        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
2835
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2836 2837 2838 2839 2840
        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
2841

Z
zhangjinchao01 已提交
2842 2843
@config_layer('lstm_step')
class LstmStepLayer(LayerBase):
Q
qijun 已提交
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
    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
2854 2855 2856
        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
2857 2858 2859 2860 2861
        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
2862 2863 2864
        self.config.active_state_type = active_state_type
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2865

Z
zhangjinchao01 已提交
2866 2867 2868
# get the specific output from the input layer.
@config_layer('get_output')
class GetOutputLayer(LayerBase):
Q
qijun 已提交
2869 2870 2871 2872
    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
2873 2874 2875 2876
        inputs = self.inputs[0]
        config_assert(inputs.input_layer_argument,
                      'input_layer_argument cannot be empty')

Q
qijun 已提交
2877

Z
zhangjinchao01 已提交
2878 2879
@config_layer('mdlstmemory')
class MDLstmLayer(LayerBase):
Q
qijun 已提交
2880 2881 2882 2883 2884 2885 2886 2887
    def __init__(self,
                 name,
                 inputs,
                 directions=True,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2888 2889
        super(MDLstmLayer, self).__init__(name, 'mdlstmemory', 0, inputs,
                                          **xargs)
Z
zhangjinchao01 已提交
2890 2891 2892 2893
        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
2894 2895
        config_assert(input_layer.size % (3 + dim_num) == 0,
                      "size % (dim_num) should be 0!")
Q
qijun 已提交
2896
        size = input_layer.size / (3 + dim_num)
Z
zhangjinchao01 已提交
2897
        self.set_layer_size(size)
Q
qijun 已提交
2898
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2899 2900 2901
        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
2902 2903
        self.create_input_parameter(0, size * size * (3 + dim_num),
                                    [size, size, 3 + dim_num])
Z
zhangjinchao01 已提交
2904
        #bias includes 3 kinds of peephole, 3+dim_num+2+dim_num
Q
qijun 已提交
2905 2906
        self.create_bias_parameter(bias, size * (5 + 2 * dim_num))

Z
zhangjinchao01 已提交
2907 2908 2909

@config_layer('gated_recurrent')
class GatedRecurrentLayer(LayerBase):
Q
qijun 已提交
2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920
    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
2921 2922 2923 2924 2925 2926
        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
2927
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2928 2929 2930
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2931

Z
zhangjinchao01 已提交
2932 2933
@config_layer('gru_step')
class GruStepLayer(LayerBase):
Q
qijun 已提交
2934 2935 2936 2937 2938 2939 2940
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_gate_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2941 2942
        super(GruStepLayer, self).__init__(name, 'gru_step', size, inputs,
                                           **xargs)
Z
zhangjinchao01 已提交
2943 2944 2945
        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
2946 2947 2948 2949 2950
        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
2951 2952 2953
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2954

Z
zhangjinchao01 已提交
2955 2956 2957 2958 2959 2960 2961
'''
 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
2962 2963


Z
zhangjinchao01 已提交
2964 2965
@config_layer('crf')
class CRFLayer(LayerBase):
Q
qijun 已提交
2966
    def __init__(self, name, size, inputs, coeff=1.0, device=None):
Z
zhangjinchao01 已提交
2967
        super(CRFLayer, self).__init__(name, 'crf', size, inputs, device=device)
Q
qijun 已提交
2968 2969
        config_assert(2 <= len(self.inputs) <= 3,
                      'CRFLayer must have 2 or 3 inputs')
Z
zhangjinchao01 已提交
2970 2971 2972
        self.create_input_parameter(0, size * (size + 2), [size, size + 2])
        self.config.coeff = coeff

Q
qijun 已提交
2973

Z
zhangjinchao01 已提交
2974 2975 2976 2977 2978 2979 2980 2981
'''
 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
2982 2983


Z
zhangjinchao01 已提交
2984 2985
@config_layer('crf_decoding')
class CRFDecodingLayer(LayerBase):
Q
qijun 已提交
2986
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
2987 2988 2989 2990 2991 2992 2993
        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
2994

Z
zhangjinchao01 已提交
2995 2996
@config_layer('ctc')
class CTCLayer(LayerBase):
Q
qijun 已提交
2997
    def __init__(self, name, size, inputs, norm_by_times=False, device=None):
Z
zhangjinchao01 已提交
2998 2999 3000 3001
        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
3002

3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023
@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
3024 3025
@config_layer('recurrent_layer_group')
class RecurrentLayerGroup(LayerBase):
Q
qijun 已提交
3026
    def __init__(self, name, device=None):
L
Luo Tao 已提交
3027 3028
        global g_pass_height_width
        g_pass_height_width = False
Z
zhangjinchao01 已提交
3029 3030 3031 3032 3033 3034
        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
3035
def Layer(name, type, **xargs):
Z
zhangjinchao01 已提交
3036 3037 3038 3039
    layers = {}
    layers.update(g_cost_map)
    layers.update(g_layer_type_map)
    layer_func = layers.get(type)
Q
qijun 已提交
3040
    config_assert(layer_func, "layer type '%s' not supported." % type)
X
xuwei06 已提交
3041
    return layer_func(name, **xargs)
Z
zhangjinchao01 已提交
3042

Q
qijun 已提交
3043

Z
zhangjinchao01 已提交
3044
@config_func
Q
qijun 已提交
3045
def ParameterHook(type, **kwargs):
Z
zhangjinchao01 已提交
3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057
    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
3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079
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
3080 3081 3082 3083 3084 3085 3086

    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
3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
    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
3098 3099
    config_assert(not momentum or not decay_rate_l1,
                  "momentum and decay_rate_l1 cannot both be non-zero")
3100 3101 3102 3103 3104

    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
3105 3106 3107 3108
    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)
3109

Q
qijun 已提交
3110 3111
    num_batches_regularization = default(num_batches_regularization,
                                         g_default_num_batches_regularization)
3112 3113 3114
    if num_batches_regularization is not None:
        para.num_batches_regularization = int(num_batches_regularization)

Z
zhangjinchao01 已提交
3115 3116 3117 3118 3119 3120
    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
3121 3122
    gradient_clipping_threshold = default(gradient_clipping_threshold,
                                          g_default_gradient_clipping_threshold)
3123 3124
    if gradient_clipping_threshold is not None:
        para.gradient_clipping_threshold = gradient_clipping_threshold
Q
qijun 已提交
3125 3126
    para.initial_strategy = default(initial_strategy,
                                    g_default_initial_strategy)
Z
zhangjinchao01 已提交
3127 3128 3129 3130 3131 3132
    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
3133 3134 3135
            print(
                "Use initial_smart, but dims not set. Initial_smart may not be used in this layer"
            )
Z
zhangjinchao01 已提交
3136 3137 3138 3139
            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)
3140 3141 3142 3143 3144 3145 3146

    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
3147 3148 3149 3150
    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")
3151 3152
    if is_shared is not None:
        para.is_shared = is_shared
Z
zhangjinchao01 已提交
3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173

    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
3174

Z
zhangjinchao01 已提交
3175 3176 3177 3178 3179
@config_func
def default_initial_mean(val):
    global g_default_initial_mean
    g_default_initial_mean = val

Q
qijun 已提交
3180

Z
zhangjinchao01 已提交
3181 3182 3183 3184 3185
@config_func
def default_initial_strategy(val):
    global g_default_initial_strategy
    g_default_initial_strategy = val

Q
qijun 已提交
3186

Z
zhangjinchao01 已提交
3187 3188 3189 3190 3191
@config_func
def default_initial_smart(val):
    global g_default_initial_smart
    g_default_initial_smart = val

Q
qijun 已提交
3192

Z
zhangjinchao01 已提交
3193 3194 3195 3196 3197
@config_func
def default_momentum(val):
    global g_default_momentum
    g_default_momentum = val

Q
qijun 已提交
3198

Z
zhangjinchao01 已提交
3199 3200 3201 3202 3203
@config_func
def default_decay_rate(val):
    global g_default_decay_rate
    g_default_decay_rate = val

Q
qijun 已提交
3204

Z
zhangjinchao01 已提交
3205 3206 3207 3208 3209
@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
3210

Z
zhangjinchao01 已提交
3211 3212 3213 3214 3215
@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
3216

Z
zhangjinchao01 已提交
3217 3218 3219 3220 3221
@config_func
def default_device(val):
    global g_default_device
    g_default_device = val

Q
qijun 已提交
3222

Z
zhangjinchao01 已提交
3223 3224 3225 3226 3227
@config_func
def default_update_hooks(val):
    global g_default_update_hooks
    g_default_update_hooks = val

Q
qijun 已提交
3228

Z
zhangjinchao01 已提交
3229 3230 3231 3232 3233
@config_func
def default_compact_func(val):
    global g_default_compact_func
    g_default_compact_func = val

Q
qijun 已提交
3234

Z
zhangjinchao01 已提交
3235 3236 3237 3238 3239
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
3240 3241 3242
        execfile(config_file,
                 make_config_environment(config_file, config_args), local_args)

Z
zhangjinchao01 已提交
3243 3244
    return Import

Q
qijun 已提交
3245

Z
zhangjinchao01 已提交
3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273
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
3274 3275 3276
    adam_beta1=0.9,
    adam_beta2=0.999,
    adam_epsilon=1e-8, )
Z
zhangjinchao01 已提交
3277

Q
qijun 已提交
3278
settings_deprecated = dict(usage_ratio=1., )
Z
zhangjinchao01 已提交
3279 3280 3281 3282

trainer_settings = dict(
    save_dir="./output/model",
    init_model_path=None,
Q
qijun 已提交
3283 3284
    start_pass=0, )

Z
zhangjinchao01 已提交
3285 3286 3287 3288 3289

@config_func
def Settings(**args):
    for k, v in args.iteritems():
        if k == "usage_ratio":
Q
qijun 已提交
3290 3291
            logger.warning(
                "Deprecated: define usage_ratio in DataConfig instead")
Z
zhangjinchao01 已提交
3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302
            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
3303

Z
zhangjinchao01 已提交
3304 3305 3306 3307
@config_func
def cluster_config(**args):
    pass

Q
qijun 已提交
3308

Z
zhangjinchao01 已提交
3309 3310 3311 3312 3313 3314 3315 3316 3317
@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
3318

Z
zhangjinchao01 已提交
3319 3320 3321 3322
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
3323

Z
zhangjinchao01 已提交
3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338
        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
3339
        get_config_arg=make_get_config_arg(config_args), )
Z
zhangjinchao01 已提交
3340 3341 3342 3343 3344

    funcs.update(g_extended_config_funcs)

    return funcs

Q
qijun 已提交
3345

Z
zhangjinchao01 已提交
3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361
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
3362

Z
zhangjinchao01 已提交
3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
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
3375

Z
zhangjinchao01 已提交
3376 3377 3378 3379
def my_fatal(s):
    logger.critical(s)
    raise Exception()

Y
Yu Yang 已提交
3380

3381
_parse_config_hooks = set()
Y
Yu Yang 已提交
3382 3383


3384 3385 3386 3387 3388 3389 3390
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
3391

Y
Yu Yang 已提交
3392

Z
zhangjinchao01 已提交
3393 3394 3395 3396 3397 3398
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()
3399 3400
    for hook in _parse_config_hooks:
        hook()
Z
zhangjinchao01 已提交
3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427

    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
3428 3429 3430 3431 3432 3433 3434 3435 3436 3437
    # 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
3438 3439 3440
        config_file.func_globals.update(
            make_config_environment("", config_args))
        config_file()
H
hanchao 已提交
3441
    else:
L
Luo Tao 已提交
3442
        execfile(config_file, make_config_environment(config_file, config_args))
Z
zhangjinchao01 已提交
3443 3444 3445
    for k, v in settings.iteritems():
        if v is None:
            continue
Q
qijun 已提交
3446
        g_config.opt_config.__setattr__(k, v)
Z
zhangjinchao01 已提交
3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472

    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
3473

Z
zhangjinchao01 已提交
3474 3475 3476 3477 3478 3479 3480 3481
if __name__ == '__main__':
    try:
        config = parse_config(sys.argv[1], '')
        config.SerializeToString()
        __real_print__(str(config))
    except:
        traceback.print_exc()
        raise