config_parser.py 124.6 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,
D
dangqingqing 已提交
496
            pad=None,
Z
zhangjinchao01 已提交
497 498 499 500 501
            format=None,
            nnz=None,
            is_static=None,
            is_shared=None,
            update_hooks=None,
502
            input_layer_argument=None,
D
dangqingqing 已提交
503 504 505 506 507
            make_layer_name_in_submodel=True, ):
        """
        @param make_layer_name_in_submodel True by defalut, you might need to
        set it carefully when adding Input in config_parser.py.
        """
Z
zhangjinchao01 已提交
508
        self.add_keys(locals())
D
dangqingqing 已提交
509 510 511
        self.input_layer_name = MakeLayerNameInSubmodel(
            input_layer_name
        ) if make_layer_name_in_submodel else input_layer_name
Z
zhangjinchao01 已提交
512

Q
qijun 已提交
513

Z
zhangjinchao01 已提交
514 515 516
# Define a projection for iexed layer
@config_class
class Projection(Input):
Q
qijun 已提交
517 518
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
519 520 521
    def __init__(
            self,
            input_layer_name,
Q
qijun 已提交
522
            size=0,  # projection output size
Z
zhangjinchao01 已提交
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
            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
542
            input_layer_argument=None, ):
Z
zhangjinchao01 已提交
543 544 545 546 547 548 549 550 551 552 553 554 555
        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
556

Z
zhangjinchao01 已提交
557 558
    def calc_parameter_size(self, input_size, output_size):
        raise NotimplementedError
Q
qijun 已提交
559

Z
zhangjinchao01 已提交
560 561 562 563 564 565 566 567 568 569
    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
570

Z
zhangjinchao01 已提交
571 572
    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
573

Z
zhangjinchao01 已提交
574 575 576
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
577

Z
zhangjinchao01 已提交
578 579 580 581 582 583
# 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
584 585 586
    def __init__(self, input_layer_name, offset, **xargs):
        super(IdentityOffsetProjection, self).__init__(input_layer_name,
                                                       **xargs)
Z
zhangjinchao01 已提交
587 588 589 590
        self.proj_conf.offset = offset

    def calc_parameter_size(self, input_size, output_size):
        return 0
Q
qijun 已提交
591

Z
zhangjinchao01 已提交
592 593 594
    def calc_parameter_dims(self, input_size, output_size):
        return []

Q
qijun 已提交
595

Z
zhangjinchao01 已提交
596 597 598 599 600 601 602
# 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
603

Z
zhangjinchao01 已提交
604 605
    def calc_parameter_size(self, input_size, output_size):
        return output_size
Q
qijun 已提交
606

Z
zhangjinchao01 已提交
607 608 609
    def calc_parameter_dims(self, input_size, output_size):
        return [1, output_size]

L
Luo Tao 已提交
610

X
xuwei06 已提交
611 612 613 614 615 616 617 618 619 620 621 622 623 624
# 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
625

Z
zhangjinchao01 已提交
626 627 628 629 630 631
@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
632

Z
zhangjinchao01 已提交
633 634 635
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
636

Z
zhangjinchao01 已提交
637 638 639 640 641 642
@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
643

Z
zhangjinchao01 已提交
644 645 646
    def calc_parameter_dims(self, input_size, output_size):
        return [input_size, output_size]

Q
qijun 已提交
647

Z
zhangjinchao01 已提交
648 649 650 651 652 653
@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
654

Z
zhangjinchao01 已提交
655 656 657
    def calc_parameter_dims(self, input_size, output_size):
        return [output_size, input_size]

Q
qijun 已提交
658

Z
zhangjinchao01 已提交
659 660 661 662
@config_class
class ContextProjection(Projection):
    type = 'context'

Q
qijun 已提交
663 664
    def __init__(self, input_layer_name, context_start, context_length,
                 trainable_padding, **xargs):
Z
zhangjinchao01 已提交
665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
        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


688 689 690 691
@config_class
class ConvProjection(Projection):
    type = 'conv'

Q
qijun 已提交
692 693 694 695 696
    def __init__(self,
                 input_layer_name,
                 num_filters=None,
                 conv_conf=None,
                 **xargs):
697 698 699 700 701
        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
702
        parse_conv(conv_conf, input_layer_name, self.proj_conf.conv_conf,
703
                   num_filters)
L
Luo Tao 已提交
704 705 706
        self.proj_conf.output_size = self.proj_conf.conv_conf.output_x * \
                                     self.proj_conf.conv_conf.output_y * \
                                     num_filters
707 708 709 710 711 712 713 714 715

    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
716 717
        gr = self.proj_conf.conv_conf.groups
        return co * ci * fh * fw / gr
718 719 720 721 722 723 724

    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
725

Z
zhangjinchao01 已提交
726 727 728
# Define a operator for mixed layer
@config_class
class Operator(Cfg):
Q
qijun 已提交
729 730
    type = None  # subclass should set it correctly

Z
zhangjinchao01 已提交
731 732
    def __init__(
            self,
Q
qijun 已提交
733
            input_layer_names, ):
Z
zhangjinchao01 已提交
734 735 736 737 738 739 740 741 742 743
        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
744

Z
zhangjinchao01 已提交
745 746 747
@config_class
class DotMulOperator(Operator):
    type = 'dot_mul'
Q
qijun 已提交
748 749 750

    def __init__(self, input_layer_names, scale=None, **xargs):
        super(DotMulOperator, self).__init__(input_layer_names, **xargs)
Z
zhangjinchao01 已提交
751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
        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
769 770 771 772 773 774 775

    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
776 777 778
        if num_filters is not None:
            self.operator_conf.num_filters = num_filters

779 780
        parse_conv(conv_conf,
                   MakeLayerNameInSubmodel(input_layer_names[0]),
Q
qijun 已提交
781
                   self.operator_conf.conv_conf, num_filters)
L
Luo Tao 已提交
782 783 784
        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
785 786 787

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

788 789
    def calc_output_size(self, input_sizes):
        return self.operator_conf.output_size
Z
zhangjinchao01 已提交
790 791 792 793 794


# please refer to the comments in proto/ModelConfig.proto
@config_class
class Conv(Cfg):
Q
qijun 已提交
795 796 797 798 799 800 801 802 803 804 805 806 807
    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
808 809
        self.add_keys(locals())
        if filter_size_y is None:
Q
qijun 已提交
810
            self.filter_size_y = filter_size
Z
zhangjinchao01 已提交
811
        if padding_y is None:
Q
qijun 已提交
812
            self.padding_y = padding
Z
zhangjinchao01 已提交
813
        if stride_y is None:
Q
qijun 已提交
814
            self.stride_y = stride
Z
zhangjinchao01 已提交
815
        if output_x is not None:
Q
qijun 已提交
816 817
            config_assert(output_x <= 0)

Z
zhangjinchao01 已提交
818

L
liaogang 已提交
819 820
@config_class
class BilinearInterp(Cfg):
L
Luo Tao 已提交
821
    def __init__(self, out_size_x=None, out_size_y=None, channels=None):
L
liaogang 已提交
822 823
        self.add_keys(locals())

Q
qijun 已提交
824

Z
zhangjinchao01 已提交
825 826
@config_class
class Pool(Cfg):
D
dangqingqing 已提交
827 828 829 830 831 832 833 834 835 836 837 838
    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
839
        self.add_keys(locals())
Q
qijun 已提交
840 841


Q
qijun 已提交
842
@config_class
Q
qijun 已提交
843
class SpatialPyramidPool(Cfg):
L
Luo Tao 已提交
844
    def __init__(self, pool_type, pyramid_height, channels):
Q
qijun 已提交
845
        self.add_keys(locals())
Z
zhangjinchao01 已提交
846

Q
qijun 已提交
847

D
dangqingqing 已提交
848 849 850 851 852 853
@config_class
class Pad(Cfg):
    def __init__(self, channels, pad_c, pad_h, pad_w):
        self.add_keys(locals())


Z
zhangjinchao01 已提交
854 855
@config_class
class Norm(Cfg):
Q
qijun 已提交
856 857 858 859 860 861 862 863 864
    def __init__(self,
                 norm_type,
                 channels,
                 size,
                 scale,
                 pow,
                 output_x=None,
                 img_size=None,
                 blocked=None):
Z
zhangjinchao01 已提交
865 866
        self.add_keys(locals())

Q
qijun 已提交
867

Z
zhangjinchao01 已提交
868 869
@config_class
class Image(Cfg):
Q
qijun 已提交
870
    def __init__(self, channels, img_size=None):
Z
zhangjinchao01 已提交
871 872
        self.add_keys(locals())

Q
qijun 已提交
873

Z
zhangjinchao01 已提交
874 875
@config_class
class BlockExpand(Cfg):
Q
qijun 已提交
876 877 878 879 880 881 882 883 884 885 886 887
    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
888 889
        self.add_keys(locals())

Q
qijun 已提交
890

891 892
@config_class
class MaxOut(Cfg):
Q
qijun 已提交
893
    def __init__(self, channels, groups, img_size_x=0, img_size_y=0):
894 895
        self.add_keys(locals())

Q
qijun 已提交
896

Z
zhangjinchao01 已提交
897 898 899 900 901 902 903 904 905 906 907 908 909
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
910 911
    data_config.data_ratio = data_ratio
    data_config.is_main_data = is_main_data
Z
zhangjinchao01 已提交
912

Q
qijun 已提交
913
    usage_ratio = default(usage_ratio, settings_deprecated["usage_ratio"])
Z
zhangjinchao01 已提交
914 915 916 917 918 919
    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
920

Z
zhangjinchao01 已提交
921
@config_func
Q
qijun 已提交
922 923 924 925 926
def SimpleData(files=None,
               feat_dim=None,
               context_len=None,
               buffer_capacity=None,
               **xargs):
Z
zhangjinchao01 已提交
927 928 929 930 931 932 933 934 935 936
    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
937

Z
zhangjinchao01 已提交
938
@config_func
Q
qijun 已提交
939 940 941 942 943 944 945 946 947 948
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
949 950 951
    data_config = DataBase(**xargs)
    data_config.type = 'py'
    if load_data_module in g_py_module_name_list:
Q
qijun 已提交
952

Z
zhangjinchao01 已提交
953 954 955
        def get_path(module):
            m = __import__(load_data_module)
            return os.path.split(os.path.realpath(m.__file__))[0]
Q
qijun 已提交
956

Z
zhangjinchao01 已提交
957 958 959
        # 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
960 961
        module_new_name = "%s_copy_%d" % (load_data_module,
                                          len(g_py_module_name_list))
Z
zhangjinchao01 已提交
962
        g_py_module_name_list.append(module_new_name)
Q
qijun 已提交
963 964 965 966
        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
967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
        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
991

Z
zhangjinchao01 已提交
992
@config_func
Q
qijun 已提交
993 994 995 996 997 998 999
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
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
    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
1020

Z
zhangjinchao01 已提交
1021 1022
#real data for training is actually provided by "sub_data" data providers.
@config_func
Q
qijun 已提交
1023
def MultiData(sub_data=[]):
Z
zhangjinchao01 已提交
1024 1025 1026 1027 1028
    data_config = DataConfig()
    data_config.type = 'multi'
    data_config.sub_data_configs.extend(sub_data)
    return data_config

Q
qijun 已提交
1029

Z
zhangjinchao01 已提交
1030
@config_func
Q
qijun 已提交
1031 1032 1033 1034 1035 1036 1037
def Data(type,
         files=None,
         feat_dim=None,
         slot_dims=None,
         context_len=None,
         buffer_capacity=None,
         **xargs):
Z
zhangjinchao01 已提交
1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072

    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
1073

L
Luo Tao 已提交
1074 1075
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1076 1077 1078 1079 1080 1081 1082
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
1083

1084
#calcualte image_size based on output_size for de-convolution (ConvTransLayer).
L
Luo Tao 已提交
1085
#It is the reverse function of cnn_output_size
1086
def cnn_image_size(output_size, filter_size, padding, stride, caffe_mode):
L
Luo Tao 已提交
1087 1088 1089
    img_size = (output_size - 1) * stride + filter_size - 2 * padding
    if not caffe_mode:
        img_size = img_size + 1
1090 1091
    return img_size

Q
qijun 已提交
1092

L
Luo Tao 已提交
1093
def get_img_size(input_layer_name, channels):
L
Luo Tao 已提交
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111
    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
1112 1113
def parse_pool(pool, input_layer_name, pool_conf):
    pool_conf.pool_type = pool.pool_type
Q
qijun 已提交
1114 1115 1116
    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
1117
                  "['max-projection', 'avg-projection', "
Q
qijun 已提交
1118
                  "'cudnn-max-pool', 'cudnn-avg-pool']" % pool.pool_type)
Z
zhangjinchao01 已提交
1119 1120 1121 1122 1123 1124

    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
1125
    pool_conf.stride_y = default(pool.stride_y, pool_conf.stride)
Z
zhangjinchao01 已提交
1126

L
Luo Tao 已提交
1127
    pool_conf.img_size, pool_conf.img_size_y = \
L
Luo Tao 已提交
1128
        get_img_size(input_layer_name, pool.channels)
Z
zhangjinchao01 已提交
1129

1130
    config_assert(not pool.start, "start is deprecated in pooling.")
Z
zhangjinchao01 已提交
1131

1132
    if pool.padding is not None:
Z
zhangjinchao01 已提交
1133
        pool_conf.padding = pool.padding
1134
    pool_conf.padding_y = default(pool.padding_y, pool_conf.padding)
D
dangqingqing 已提交
1135 1136 1137 1138 1139 1140
    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
1141

Z
zhangjinchao01 已提交
1142

Q
qijun 已提交
1143
def parse_spp(spp, input_layer_name, spp_conf):
L
Luo Tao 已提交
1144
    parse_image(spp, input_layer_name, spp_conf.image_conf)
Q
qijun 已提交
1145 1146
    spp_conf.pool_type = spp.pool_type
    config_assert(spp.pool_type in ['max-projection', 'avg-projection'],
Q
qijun 已提交
1147 1148
                  "pool-type %s is not in "
                  "['max-projection', 'avg-projection']" % spp.pool_type)
Q
qijun 已提交
1149
    spp_conf.pyramid_height = spp.pyramid_height
Q
qijun 已提交
1150

Q
qijun 已提交
1151

Z
zhangjinchao01 已提交
1152 1153
def parse_image(image, input_layer_name, image_conf):
    image_conf.channels = image.channels
L
Luo Tao 已提交
1154
    image_conf.img_size, image_conf.img_size_y = \
L
Luo Tao 已提交
1155
        get_img_size(input_layer_name, image_conf.channels)
Q
qijun 已提交
1156

Z
zhangjinchao01 已提交
1157 1158 1159 1160

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
1161 1162
                  "norm-type %s is not in [rnorm, 'cmrnorm-projection']" %
                  norm.norm_type)
Z
zhangjinchao01 已提交
1163 1164 1165 1166 1167 1168
    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
1169
    norm_conf.img_size, norm_conf.img_size_y = \
L
Luo Tao 已提交
1170
        get_img_size(input_layer_name, norm.channels)
Z
zhangjinchao01 已提交
1171
    norm_conf.output_x = norm_conf.img_size
L
Luo Tao 已提交
1172
    norm_conf.output_y = norm_conf.img_size_y
Z
zhangjinchao01 已提交
1173 1174 1175
    if norm.norm_type in ['cmrnorm-projection']:
        norm_conf.scale /= norm.size
    else:
Q
qijun 已提交
1176 1177
        norm_conf.scale /= norm.size**2

1178

L
Luo Tao 已提交
1179 1180
#caffe_mode: compute the output size using floor instead of ceil,
#            which is consistent of caffe and CuDNN's convention.
1181
def parse_conv(conv, input_layer_name, conv_conf, num_filters, trans=False):
Z
zhangjinchao01 已提交
1182 1183 1184 1185 1186 1187 1188 1189 1190
    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
1191

1192
    if not trans:
1193
        conv_conf.filter_channels = conv.channels / conv.groups
L
Luo Tao 已提交
1194
        conv_conf.img_size, conv_conf.img_size_y = \
L
Luo Tao 已提交
1195
            get_img_size(input_layer_name, conv.channels)
1196
        conv_conf.output_x = cnn_output_size(
Q
qijun 已提交
1197 1198
            conv_conf.img_size, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1199 1200 1201
        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)
1202
    else:
1203
        conv_conf.filter_channels = num_filters / conv.groups
L
Luo Tao 已提交
1204
        conv_conf.output_x, conv_conf.output_y = \
L
Luo Tao 已提交
1205
            get_img_size(input_layer_name, conv.channels)
1206
        conv_conf.img_size = cnn_image_size(
Q
qijun 已提交
1207 1208
            conv_conf.output_x, conv_conf.filter_size, conv_conf.padding,
            conv_conf.stride, conv_conf.caffe_mode)
L
Luo Tao 已提交
1209
        conv_conf.img_size_y = cnn_image_size(
L
Luo Tao 已提交
1210 1211
            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
1212

1213

Z
zhangjinchao01 已提交
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
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:
1227
        block_expand_conf.output_x = cnn_output_size(
1228
            block_expand.img_size_x, block_expand.block_x,
1229
            block_expand.padding_x, block_expand.stride_x, False)
Z
zhangjinchao01 已提交
1230 1231

    if block_expand_conf.img_size_y == 0:
1232
        block_expand_conf.output_y = 0
Z
zhangjinchao01 已提交
1233
    else:
1234
        block_expand_conf.output_y = cnn_output_size(
1235
            block_expand.img_size_y, block_expand.block_y,
1236
            block_expand.padding_y, block_expand.stride_y, False)
Z
zhangjinchao01 已提交
1237

Q
qijun 已提交
1238

1239
def parse_maxout(maxout, input_layer_name, maxout_conf):
L
Luo Tao 已提交
1240
    parse_image(maxout, input_layer_name, maxout_conf.image_conf)
1241
    maxout_conf.groups = maxout.groups
1242

Q
qijun 已提交
1243

Z
zhangjinchao01 已提交
1244 1245 1246 1247 1248 1249
# Define an evaluator
@config_func
def Evaluator(
        name,
        type,
        inputs,
Q
qijun 已提交
1250 1251 1252 1253 1254 1255 1256
        chunk_scheme=None,
        num_chunk_types=None,
        classification_threshold=None,
        positive_label=None,
        dict_file=None,
        result_file=None,
        num_results=None,
1257 1258
        delimited=None,
        excluded_chunk_types=None, ):
Z
zhangjinchao01 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
    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)

1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
    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
1286

1287 1288 1289
    if excluded_chunk_types:
        evaluator.excluded_chunk_types.extend(excluded_chunk_types)

Q
qijun 已提交
1290

Z
zhangjinchao01 已提交
1291 1292 1293 1294 1295
class LayerBase(object):
    def __init__(
            self,
            name,
            type,
Q
qijun 已提交
1296
            size,  # size can be 0. In this case, subclass should set it.
Z
zhangjinchao01 已提交
1297 1298 1299 1300
            inputs,
            device=None,
            active_type="",
            drop_rate=0.,
1301
            coeff=None):
Z
zhangjinchao01 已提交
1302
        config_assert('@' not in name,
Q
qijun 已提交
1303
                      "layer name: %s contain special character @" % name)
Z
zhangjinchao01 已提交
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
        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()
1319
        assert isinstance(self.config, LayerConfig)
Z
zhangjinchao01 已提交
1320 1321 1322
        self.config.name = name
        self.config.type = type
        self.config.active_type = active_type
1323 1324
        if coeff is not None:
            self.config.coeff = float(coeff)
Z
zhangjinchao01 已提交
1325 1326 1327 1328 1329 1330 1331
        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
1332
        elif g_default_device is not None:
Z
zhangjinchao01 已提交
1333 1334 1335 1336 1337 1338 1339 1340 1341
            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
1342 1343
                    input_layer_name=input,
                    parameter_name=gen_parameter_name(name, input_index))
Z
zhangjinchao01 已提交
1344 1345 1346 1347 1348 1349 1350 1351
                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
1352
                self.operators.append(input)
Z
zhangjinchao01 已提交
1353 1354 1355 1356
                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
1357
                raise ValueError('Wrong type for inputs: %s' % type_of(input))
Z
zhangjinchao01 已提交
1358
            config_assert(input_layer_name in g_layer_map,
Q
qijun 已提交
1359 1360
                          "Unknown input layer '%s' for layer %s" %
                          (input_layer_name, name))
Z
zhangjinchao01 已提交
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371
            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
1372 1373 1374 1375 1376 1377
        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
1378 1379 1380 1381 1382 1383
    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
1384
            bias,  # True/False or BiasCfg
Z
zhangjinchao01 已提交
1385
            size,
Q
qijun 已提交
1386 1387 1388
            dims=None,
            for_self=True,  # whether create bias for layer self
    ):
Z
zhangjinchao01 已提交
1389 1390 1391 1392 1393 1394

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

Q
qijun 已提交
1395 1396 1397
        config_assert(
            type_of(bias) == bool or type_of(bias) == Bias,
            'Incorrect type for bias: %s' % type_of(bias))
Z
zhangjinchao01 已提交
1398 1399 1400 1401 1402 1403 1404 1405 1406

        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:
1407 1408
                assert isinstance(self.config, LayerConfig)

Z
zhangjinchao01 已提交
1409 1410 1411
                Parameter(
                    bias.parameter_name,
                    size,
Q
qijun 已提交
1412 1413
                    self.config.device
                    if self.config.HasField('device') else None,
Z
zhangjinchao01 已提交
1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
                    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
1425 1426
                    gradient_clipping_threshold=bias.
                    gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1427
                    is_static=bias.is_static,
Q
qijun 已提交
1428
                    is_shared=bias.is_shared, )
Z
zhangjinchao01 已提交
1429 1430 1431 1432 1433
            if for_self:
                self.config.bias_parameter_name = bias.parameter_name
            else:
                return bias.parameter_name

Q
qijun 已提交
1434 1435 1436 1437 1438 1439
    def create_input_parameter(self,
                               input_index,
                               size,
                               dims=None,
                               sparse=None,
                               format=None):
Z
zhangjinchao01 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453
        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
1454 1455
            config_assert(size == para.size, (
                'Shared parameter "%s" does not ' + 'have same size: %s vs. %s')
Z
zhangjinchao01 已提交
1456 1457
                          % (input_config.parameter_name, para.size, size))

Q
qijun 已提交
1458 1459
            config_assert(dims == para.dims, (
                'Shared parameter "%s" does not ' + 'have same dims: %s vs. %s')
Z
zhangjinchao01 已提交
1460 1461 1462 1463 1464 1465
                          % (input_config.parameter_name, para.dims, dims))
            return

        Parameter(
            input_config.parameter_name,
            size,
1466
            self.config.device if self.config.HasField("device") else None,
Z
zhangjinchao01 已提交
1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
            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
1479 1480
            gradient_clipping_threshold=input_config.
            gradient_clipping_threshold,
Z
zhangjinchao01 已提交
1481 1482 1483 1484
            sparse=sparse,
            format=format,
            is_static=input_config.is_static,
            is_shared=input_config.is_shared,
Q
qijun 已提交
1485
            update_hooks=input_config.update_hooks)
Z
zhangjinchao01 已提交
1486 1487 1488 1489 1490 1491 1492 1493 1494

    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
1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511
    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
1512

Z
zhangjinchao01 已提交
1513 1514
@config_layer('multi_class_cross_entropy_with_selfnorm')
class MultiClassCrossEntropySelfNormCostLayer(LayerBase):
Q
qijun 已提交
1515 1516 1517
    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
1518 1519
        self.config.softmax_selfnorm_alpha = softmax_selfnorm_alpha

Q
qijun 已提交
1520

Z
zhangjinchao01 已提交
1521 1522
@config_layer('fc')
class FCLayer(LayerBase):
Q
qijun 已提交
1523
    def __init__(self, name, size, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533
        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
1534 1535
            else:
                sparse = None
Z
zhangjinchao01 已提交
1536

Q
qijun 已提交
1537 1538
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1539 1540
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1541

Z
zhangjinchao01 已提交
1542 1543
@config_layer('selective_fc')
class SelectiveFCLayer(LayerBase):
Q
qijun 已提交
1544 1545 1546 1547 1548 1549 1550 1551 1552 1553
    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
1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
        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
1574 1575
                          ("if indices of selected columns are not specified, "
                           "selective_fc Layer has at least two inputs"))
Z
zhangjinchao01 已提交
1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
            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
1588 1589
            self.create_input_parameter(input_index, psize, dims, sparse,
                                        format)
Z
zhangjinchao01 已提交
1590 1591
        self.create_bias_parameter(bias, self.config.size)

Q
qijun 已提交
1592

1593 1594
@config_layer('print')
class PrintLayer(LayerBase):
Q
qijun 已提交
1595
    def __init__(self, name, inputs):
1596 1597
        super(PrintLayer, self).__init__(name, 'print', 0, inputs)

Q
qijun 已提交
1598

Y
yuan 已提交
1599 1600
@config_layer('priorbox')
class PriorBoxLayer(LayerBase):
G
gaoyuan 已提交
1601 1602
    def __init__(self, name, inputs, size, min_size, max_size, aspect_ratio,
                 variance):
Y
yuan 已提交
1603
        super(PriorBoxLayer, self).__init__(name, 'priorbox', 0, inputs)
G
gaoyuan 已提交
1604
        config_assert(len(inputs) == 2, 'PriorBoxLayer must have 2 inputs')
G
gaoyuan 已提交
1605 1606 1607 1608 1609 1610 1611
        input_layer = self.get_input_layer(1)
        config_assert(
            input_layer.type == 'data',
            'Expecting the second input layer of an priorbox layer to be '
            'a data layer')
        config_assert(input_layer.width > 0, 'The data layer must set width')
        config_assert(input_layer.height > 0, 'The data layer must set height')
G
gaoyuan 已提交
1612
        config_assert(len(variance) == 4, 'The variance must have 4 inputs')
Y
yuan 已提交
1613 1614 1615 1616 1617 1618
        self.config.inputs[0].priorbox_conf.min_size.extend(min_size)
        self.config.inputs[0].priorbox_conf.max_size.extend(max_size)
        self.config.inputs[0].priorbox_conf.aspect_ratio.extend(aspect_ratio)
        self.config.inputs[0].priorbox_conf.variance.extend(variance)
        self.config.size = size

Q
qijun 已提交
1619

Z
zhangjinchao01 已提交
1620 1621
@config_layer('data')
class DataLayer(LayerBase):
L
Luo Tao 已提交
1622
    def __init__(self, name, size, height=None, width=None, device=None):
Q
qijun 已提交
1623 1624
        super(DataLayer, self).__init__(
            name, 'data', size, inputs=[], device=device)
L
Luo Tao 已提交
1625 1626
        if height and width:
            self.set_layer_height_width(height, width)
Q
qijun 已提交
1627

Z
zhangjinchao01 已提交
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654

'''
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
1655 1656


Z
zhangjinchao01 已提交
1657 1658
@config_layer('data_norm')
class DataNormLayer(LayerBase):
Q
qijun 已提交
1659
    def __init__(self, name, inputs, data_norm_strategy="z-score", device=None):
Z
zhangjinchao01 已提交
1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
        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
1671

Z
zhangjinchao01 已提交
1672 1673 1674
@config_layer('prelu')
class ParameterReluLayer(LayerBase):
    layer_type = 'prelu'
Q
qijun 已提交
1675 1676

    def __init__(self, name, inputs, partial_sum=1, **args):
Z
zhangjinchao01 已提交
1677 1678 1679 1680 1681 1682 1683 1684
        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
1685

Z
zhangjinchao01 已提交
1686 1687 1688
@config_layer('conv')
class ConvLayerBase(LayerBase):
    layer_type = 'conv'
Q
qijun 已提交
1689 1690 1691 1692 1693 1694 1695 1696

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
Z
zhangjinchao01 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
        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
1713
            (parallel_nn == 0 or self.config.device > -1)):
Z
zhangjinchao01 已提交
1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
            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
1726 1727
            parse_conv(self.inputs[input_index].conv, input_layer.name,
                       conv_conf, num_filters)
Z
zhangjinchao01 已提交
1728 1729
            psize = self.calc_parameter_size(conv_conf)
            self.create_input_parameter(input_index, psize)
L
Luo Tao 已提交
1730 1731
            self.set_cnn_layer(name, conv_conf.output_y, conv_conf.output_x,
                               self.config.num_filters)
Z
zhangjinchao01 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741

        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
1742

Z
zhangjinchao01 已提交
1743 1744 1745 1746
@config_layer('exconv')
class ConvLayer(ConvLayerBase):
    layer_type = 'exconv'

Q
qijun 已提交
1747

Z
zhangjinchao01 已提交
1748 1749 1750 1751
@config_layer('cudnn_conv')
class ConvLayer(ConvLayerBase):
    layer_type = 'cudnn_conv'

1752 1753 1754 1755

@config_layer('convt')
class ConvTransLayerBase(LayerBase):
    layer_type = 'convt'
Q
qijun 已提交
1756 1757 1758 1759 1760 1761 1762 1763

    def __init__(self,
                 name,
                 inputs=[],
                 bias=True,
                 num_filters=None,
                 shared_biases=False,
                 **xargs):
1764
        super(ConvTransLayerBase, self).__init__(
1765 1766 1767 1768 1769 1770 1771 1772
            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))

1773 1774
        # cudnn_convt has not been implemented so use exconvt only
        self.layer_type = "exconvt"
1775 1776 1777 1778 1779 1780 1781 1782
        # 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)
1783
            parse_conv(
1784 1785
                self.inputs[input_index].conv,
                input_layer.name,
1786
                self.config.inputs[input_index].conv_conf,
1787
                num_filters,
1788
                trans=True)
1789 1790 1791 1792 1793
            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
1794
                (conv_conf.img_size**2) * self.config.num_filters)
1795 1796 1797 1798 1799 1800 1801

        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):
1802
        return conv_conf.channels * conv_conf.filter_channels \
1803 1804
                    * (conv_conf.filter_size * conv_conf.filter_size_y)

Q
qijun 已提交
1805

1806 1807 1808 1809
@config_layer('exconvt')
class ConvTransLayer(ConvTransLayerBase):
    layer_type = 'exconvt'

Q
qijun 已提交
1810

Z
zhangjinchao01 已提交
1811 1812
@config_layer('norm')
class NormLayer(LayerBase):
Q
qijun 已提交
1813 1814 1815
    def __init__(self, name, inputs, device=None):
        super(NormLayer, self).__init__(
            name, 'norm', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1816 1817 1818
        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
1819 1820 1821 1822
            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
1823

Z
zhangjinchao01 已提交
1824 1825 1826

@config_layer('pool')
class PoolLayer(LayerBase):
Q
qijun 已提交
1827 1828 1829
    def __init__(self, name, inputs, device=None):
        super(PoolLayer, self).__init__(
            name, 'pool', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1830 1831 1832
        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
1833 1834 1835 1836
            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
1837

Z
zhangjinchao01 已提交
1838

Q
qijun 已提交
1839 1840
@config_layer('spp')
class SpatialPyramidPoolLayer(LayerBase):
Q
qijun 已提交
1841 1842 1843
    def __init__(self, name, inputs, device=None):
        super(SpatialPyramidPoolLayer, self).__init__(
            name, 'spp', 0, inputs=inputs, device=device)
Q
qijun 已提交
1844 1845 1846
        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
1847 1848 1849
            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
1850

Q
qijun 已提交
1851

D
dangqingqing 已提交
1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
@config_layer('pad')
class PadLayer(LayerBase):
    def __init__(self, name, inputs, **xargs):
        super(PadLayer, self).__init__(name, 'pad', 0, inputs=inputs, **xargs)
        pad = self.inputs[0].pad
        self.config.inputs[0].pad_conf.pad_c.extend(pad.pad_c)
        self.config.inputs[0].pad_conf.pad_h.extend(pad.pad_h)
        self.config.inputs[0].pad_conf.pad_w.extend(pad.pad_w)

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


Z
zhangjinchao01 已提交
1871 1872 1873
@config_layer('batch_norm')
class BatchNormLayer(LayerBase):
    layer_type = 'batch_norm'
Q
qijun 已提交
1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884

    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
1885 1886 1887 1888
        if inputs is None:
            inputs = []
        elif not isinstance(inputs, list):
            inputs = [inputs]
Q
qijun 已提交
1889 1890
        config_assert(
            len(inputs) == 1, "BatchNormLayer must have one and only one input")
Z
zhangjinchao01 已提交
1891 1892 1893 1894 1895 1896 1897 1898
        # 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
1899 1900 1901 1902 1903 1904
            inputs.append(
                Input(
                    inputs[0].input_layer_name,
                    initial_std=0.0,
                    initial_mean=0.0,
                    is_static=True,
1905
                    is_shared=is_shared,
D
dangqingqing 已提交
1906
                    make_layer_name_in_submodel=False, ))
Z
zhangjinchao01 已提交
1907 1908 1909 1910 1911 1912 1913

        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 \
1914
            cudnn_version >= 4007
Z
zhangjinchao01 已提交
1915
        self.layer_type = "cudnn_batch_norm" if use_cudnn else "batch_norm"
Q
qijun 已提交
1916 1917 1918 1919 1920 1921 1922 1923
        super(BatchNormLayer, self).__init__(
            name,
            self.layer_type,
            0,
            active_type=active_type,
            inputs=inputs,
            device=device,
            **xargs)
Z
zhangjinchao01 已提交
1924 1925 1926 1927 1928 1929

        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
1930
        input_layer = self.get_input_layer(0)
Z
zhangjinchao01 已提交
1931
        image_conf = self.config.inputs[0].image_conf
L
Luo Tao 已提交
1932
        parse_image(self.inputs[0].image, input_layer.name, image_conf)
1933 1934 1935 1936 1937

        # Only pass the width and height of input to batch_norm layer 
        # when either of it is non-zero. 
        if input_layer.width != 0 or input_layer.height != 0:
            self.set_cnn_layer(name, image_conf.img_size_y, image_conf.img_size,
D
dangqingqing 已提交
1938
                               image_conf.channels, False)
1939 1940
        else:
            self.set_layer_size(input_layer.size)
Z
zhangjinchao01 已提交
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952

        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
1953

Z
zhangjinchao01 已提交
1954 1955
@config_layer('trans')
class TransLayer(LayerBase):
Q
qijun 已提交
1956 1957 1958 1959 1960 1961
    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
1962 1963
        self.set_layer_size(self.get_input_layer(0).size)

Q
qijun 已提交
1964

Z
zhangjinchao01 已提交
1965 1966
@config_layer('resize')
class ResizeLayer(LayerBase):
Q
qijun 已提交
1967 1968 1969 1970 1971 1972 1973
    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
1974 1975 1976

@config_layer('blockexpand')
class BlockExpandLayer(LayerBase):
Q
qijun 已提交
1977 1978 1979
    def __init__(self, name, inputs, device=None):
        super(BlockExpandLayer, self).__init__(
            name, 'blockexpand', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
1980 1981
        for input_index in xrange(len(self.inputs)):
            input_layer = self.get_input_layer(input_index)
Q
qijun 已提交
1982 1983
            parse_block_expand(
                self.inputs[input_index].block_expand, input_layer.name,
Z
zhangjinchao01 已提交
1984
                self.config.inputs[input_index].block_expand_conf)
Q
qijun 已提交
1985 1986 1987 1988 1989 1990
            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
1991

1992 1993
@config_layer('maxout')
class MaxOutLayer(LayerBase):
Q
qijun 已提交
1994 1995 1996
    def __init__(self, name, inputs, **xargs):
        super(MaxOutLayer, self).__init__(
            name, 'maxout', 0, inputs=inputs, **xargs)
1997 1998
        input_layer = self.get_input_layer(0)
        maxout_conf = self.config.inputs[0].maxout_conf
L
Luo Tao 已提交
1999
        parse_maxout(self.inputs[0].maxout, input_layer.name, maxout_conf)
L
Luo Tao 已提交
2000 2001 2002
        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
2003

2004

Z
zhangjinchao01 已提交
2005 2006 2007 2008
# key: cost type
# value: cost class
g_cost_map = {}

Q
qijun 已提交
2009

Z
zhangjinchao01 已提交
2010 2011 2012
# 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
2013 2014
        super(type(cls), cls).__init__(
            name, cost_type, 1, inputs, device=device, coeff=coeff)
Z
zhangjinchao01 已提交
2015

Q
qijun 已提交
2016
    cls = type(class_name, (LayerBase, ), dict(__init__=init))
Z
zhangjinchao01 已提交
2017 2018 2019
    global g_cost_map
    g_cost_map[cost_type] = cls

Q
qijun 已提交
2020

Z
zhangjinchao01 已提交
2021 2022 2023 2024 2025 2026 2027 2028
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
2029
define_cost('SumCost', 'sum_cost')
Z
zhangjinchao01 已提交
2030

Q
qijun 已提交
2031

Z
zhangjinchao01 已提交
2032 2033
@config_layer('hsigmoid')
class HierarchicalSigmoidLayer(LayerBase):
Q
qijun 已提交
2034
    def __init__(self, name, num_classes, inputs, device=None, bias=True):
Z
zhangjinchao01 已提交
2035 2036
        super(HierarchicalSigmoidLayer, self).__init__(
            name, 'hsigmoid', 1, inputs=inputs, device=device)
Q
qijun 已提交
2037 2038 2039
        config_assert(
            len(self.inputs) >= 2,
            'HierarchicalSigmoidLayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2040 2041 2042 2043 2044 2045 2046 2047
        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
2048

Z
zhangjinchao01 已提交
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
'''
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
2073 2074


Z
zhangjinchao01 已提交
2075 2076
@config_layer('lambda_cost')
class LambdaCost(LayerBase):
Q
qijun 已提交
2077
    def __init__(self, name, inputs, NDCG_num=5, max_sort_size=-1, device=None):
Z
zhangjinchao01 已提交
2078 2079
        super(LambdaCost, self).__init__(
            name, 'lambda_cost', 1, inputs=inputs, device=device)
Q
qijun 已提交
2080
        config_assert(len(self.inputs) == 2, 'lambdaCost must have 2 inputs')
Z
zhangjinchao01 已提交
2081 2082
        self.config.NDCG_num = NDCG_num
        if max_sort_size != -1:
Q
qijun 已提交
2083 2084 2085
            config_assert(
                NDCG_num <= max_sort_size,
                'NDCG_num must be less than or equal to max_sort_size')
Z
zhangjinchao01 已提交
2086 2087
        self.config.max_sort_size = max_sort_size

Q
qijun 已提交
2088

Z
zhangjinchao01 已提交
2089 2090
@config_layer('nce')
class NCELayer(LayerBase):
Q
qijun 已提交
2091 2092 2093 2094 2095 2096 2097 2098
    def __init__(self,
                 name,
                 num_classes,
                 inputs,
                 num_neg_samples=10,
                 neg_sampling_dist=None,
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2099
        super(NCELayer, self).__init__(name, 'nce', 1, inputs=inputs, **xargs)
Q
qijun 已提交
2100 2101
        config_assert(
            len(self.inputs) >= 2, 'NCELayer must have at least 2 inputs')
Z
zhangjinchao01 已提交
2102 2103
        self.config.num_classes = num_classes
        if neg_sampling_dist is not None:
Q
qijun 已提交
2104 2105 2106 2107
            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
2108
            s = sum(neg_sampling_dist)
Q
qijun 已提交
2109 2110 2111
            config_assert(
                abs(s - 1) < 1e-5,
                'The sum of neg_sampling_dist (%s) is not 1' % s)
Z
zhangjinchao01 已提交
2112 2113 2114 2115 2116

            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
2117
        input_layer = self.get_input_layer(num_real_inputs)
Z
zhangjinchao01 已提交
2118 2119 2120 2121
        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
2122 2123
        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
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136
            # 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
2137
    def __init__(self, name, inputs, bias=True, **xargs):
Z
zhangjinchao01 已提交
2138 2139
        super(AddToLayer, self).__init__(
            name, 'addto', 0, inputs=inputs, **xargs)
Q
qijun 已提交
2140
        config_assert(len(inputs) > 0, 'inputs cannot be empty for AddToLayer')
Z
zhangjinchao01 已提交
2141 2142 2143 2144 2145
        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
2146

Z
zhangjinchao01 已提交
2147 2148
@config_layer('agent')
class AgentLayer(LayerBase):
Q
qijun 已提交
2149 2150 2151 2152
    def __init__(self, name, size, device=None):
        super(AgentLayer, self).__init__(
            name, 'agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2153 2154 2155

@config_layer('sequence_agent')
class SequenceAgentLayer(LayerBase):
Q
qijun 已提交
2156
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2157 2158 2159
        super(SequenceAgentLayer, self).__init__(
            name, 'sequence_agent', size, inputs=[], device=device)

Q
qijun 已提交
2160

Z
zhangjinchao01 已提交
2161 2162
@config_layer('gather_agent')
class GatherAgentLayer(LayerBase):
Q
qijun 已提交
2163
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2164 2165 2166
        super(GatherAgentLayer, self).__init__(
            name, 'gather_agent', size, inputs=[], device=device)

Q
qijun 已提交
2167

Z
zhangjinchao01 已提交
2168 2169
@config_layer('scatter_agent')
class ScatterAgentLayer(LayerBase):
Q
qijun 已提交
2170
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2171 2172 2173
        super(ScatterAgentLayer, self).__init__(
            name, 'scatter_agent', size, inputs=[], device=device)

Q
qijun 已提交
2174

Z
zhangjinchao01 已提交
2175 2176
@config_layer('sequence_gather_agent')
class SequenceGatherAgentLayer(LayerBase):
Q
qijun 已提交
2177
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2178
        super(SequenceGatherAgentLayer, self).__init__(
Q
qijun 已提交
2179 2180
            name, 'sequence_gather_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2181 2182 2183

@config_layer('sequence_scatter_agent')
class SequenceScatterAgentLayer(LayerBase):
Q
qijun 已提交
2184
    def __init__(self, name, size, device=None):
Z
zhangjinchao01 已提交
2185
        super(SequenceScatterAgentLayer, self).__init__(
Q
qijun 已提交
2186 2187
            name, 'sequence_scatter_agent', size, inputs=[], device=device)

Z
zhangjinchao01 已提交
2188 2189 2190

@config_layer('multiplex')
class MultiplexLayer(LayerBase):
Q
qijun 已提交
2191 2192 2193 2194 2195
    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
2196
        for i in range(1, len(inputs)):
Q
qijun 已提交
2197 2198 2199 2200 2201
            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
2202 2203

@config_func
Q
qijun 已提交
2204 2205 2206
def Link(
        name,
        has_subseq=False, ):
Z
zhangjinchao01 已提交
2207 2208 2209 2210 2211
    link_config = LinkConfig()
    link_config.link_name = name
    link_config.has_subseq = has_subseq
    return link_config

Q
qijun 已提交
2212

Z
zhangjinchao01 已提交
2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226
# 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
2227 2228 2229 2230 2231 2232 2233 2234
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
2235 2236 2237 2238 2239 2240
    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
2241
                  'Memory should be used in recurrent layer group only')
Z
zhangjinchao01 已提交
2242 2243 2244 2245
    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
2246
    options = sum((boot_layer is not None, bool(boot_bias),
Z
zhangjinchao01 已提交
2247
                   boot_with_const_id is not None))
Q
qijun 已提交
2248 2249 2250 2251
    config_assert(
        options <= 1,
        'take one option at most from boot_layer, boot_bias, or boot_with_const_id'
    )
Z
zhangjinchao01 已提交
2252 2253 2254
    if boot_layer is not None:
        boot_layer = MakeLayerNameInParentSubmodel(boot_layer)
        config_assert(boot_layer in g_layer_map,
Q
qijun 已提交
2255 2256
                      'boot_layer "%s" does not correspond to a layer name' %
                      boot_layer)
Z
zhangjinchao01 已提交
2257 2258 2259
        memory.boot_layer_name = boot_layer
    elif boot_bias:
        memory.boot_bias_parameter_name = agent_layer.create_bias_parameter(
Q
qijun 已提交
2260
            boot_bias, size, for_self=False)
Z
zhangjinchao01 已提交
2261 2262 2263 2264 2265
        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
2266

Z
zhangjinchao01 已提交
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
# 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
2278 2279 2280 2281
        eos_layer_name="eos_check",
        num_results_per_sample=1,
        beam_size=1,
        log_prob=None, ):
Z
zhangjinchao01 已提交
2282 2283 2284 2285 2286 2287 2288 2289 2290
    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
2291

Z
zhangjinchao01 已提交
2292 2293
@config_layer('expand')
class ExpandLayer(LayerBase):
Q
qijun 已提交
2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
    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
2310 2311 2312

@config_layer('featmap_expand')
class FeatMapExpandLayer(LayerBase):
Q
qijun 已提交
2313 2314 2315 2316 2317 2318
    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
2319
            self.config.num_filters = num_filters
Q
qijun 已提交
2320
        else:
Z
zhangjinchao01 已提交
2321
            logger.fatal("FeatMapExpandLayer must specify num_filters.")
Q
qijun 已提交
2322
        self.set_layer_size(self.get_input_layer(0).size * num_filters)
Z
zhangjinchao01 已提交
2323 2324 2325 2326


@config_layer('max')
class MaxLayer(LayerBase):
Q
qijun 已提交
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336
    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
2337
        config_assert(len(self.inputs) == 1, 'MaxLayer must have 1 input')
Q
qijun 已提交
2338 2339
        self.config.trans_type = trans_type
        self.config.active_type = active_type
Z
zhangjinchao01 已提交
2340 2341 2342 2343
        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)
2344 2345
        if output_max_index is not None:
            self.config.output_max_index = output_max_index
Z
zhangjinchao01 已提交
2346 2347 2348 2349


@config_layer('maxid')
class MaxIdLayer(LayerBase):
Q
qijun 已提交
2350
    def __init__(self, name, inputs, beam_size=None, device=None):
Z
zhangjinchao01 已提交
2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367
        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
2368
    def __init__(self, name, inputs, eos_id, device=None):
Z
zhangjinchao01 已提交
2369 2370 2371
        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
2372
        self.set_layer_size(2)  # boolean output
Z
zhangjinchao01 已提交
2373 2374
        self.config.eos_id = eos_id

Q
qijun 已提交
2375

Z
zhangjinchao01 已提交
2376 2377
@config_layer('seqlastins')
class SequenceLastInstanceLayer(LayerBase):
Q
qijun 已提交
2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394
    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
2395 2396 2397 2398 2399
        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
2400

Z
zhangjinchao01 已提交
2401 2402 2403 2404 2405 2406 2407 2408 2409
@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
2410 2411 2412 2413 2414 2415 2416 2417
            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
2418 2419
        self.config.select_first = True

Q
qijun 已提交
2420

Z
zhangjinchao01 已提交
2421 2422
@config_layer('seqconcat')
class SequenceConcatLayer(LayerBase):
Q
qijun 已提交
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
    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
2438 2439 2440 2441 2442
        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
2443

Z
zhangjinchao01 已提交
2444 2445
@config_layer('seqreshape')
class SequenceReshapeLayer(LayerBase):
Q
qijun 已提交
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_type='linear',
                 device=None,
                 bias=False):
        super(SequenceReshapeLayer, self).__init__(
            name,
            'seqreshape',
Z
zhangjinchao01 已提交
2456
            size,
Q
qijun 已提交
2457 2458 2459 2460 2461
            inputs=inputs,
            device=device,
            active_type=active_type)
        config_assert(
            len(inputs) == 1, 'SequenceReshapeLayer must have 1 inputs')
Z
zhangjinchao01 已提交
2462 2463 2464
        self.set_layer_size(size)
        self.create_bias_parameter(bias, size)

Q
qijun 已提交
2465

Z
zhangjinchao01 已提交
2466 2467
@config_layer('subseq')
class SubSequenceLayer(LayerBase):
Q
qijun 已提交
2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480
    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
2481 2482 2483 2484 2485 2486
        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
2487

Z
zhangjinchao01 已提交
2488 2489
@config_layer('out_prod')
class OuterProdLayer(LayerBase):
Q
qijun 已提交
2490 2491 2492
    def __init__(self, name, inputs, device=None):
        super(OuterProdLayer, self).__init__(
            name, 'out_prod', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2493 2494 2495 2496 2497
        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
2498

Z
zhangjinchao01 已提交
2499 2500
@config_layer('power')
class PowerLayer(LayerBase):
Q
qijun 已提交
2501 2502 2503
    def __init__(self, name, inputs, device=None):
        super(PowerLayer, self).__init__(
            name, 'power', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2504 2505 2506 2507
        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
2508 2509 2510
        config_assert(1 == input_layer0.size,
                      'The left input is the exponent and should be of size 1')

Z
zhangjinchao01 已提交
2511 2512 2513

@config_layer('slope_intercept')
class SlopeInterceptLayer(LayerBase):
Q
qijun 已提交
2514 2515 2516
    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
2517 2518 2519 2520 2521 2522
        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
2523

Z
zhangjinchao01 已提交
2524 2525
@config_layer('scaling')
class ScalingLayer(LayerBase):
Q
qijun 已提交
2526 2527 2528
    def __init__(self, name, inputs, device=None):
        super(ScalingLayer, self).__init__(
            name, 'scaling', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2529 2530 2531 2532
        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
2533 2534 2535
        config_assert(1 == input_layer0.size,
                      'The left input should be of size 1')

Z
zhangjinchao01 已提交
2536 2537 2538

@config_layer('conv_shift')
class ConvShiftLayer(LayerBase):
Q
qijun 已提交
2539 2540 2541
    def __init__(self, name, inputs, device=None):
        super(ConvShiftLayer, self).__init__(
            name, 'conv_shift', 0, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2542 2543 2544 2545
        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
2546

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

Q
qijun 已提交
2559

Z
zhangjinchao01 已提交
2560 2561
@config_layer('interpolation')
class InterpolationLayer(LayerBase):
Q
qijun 已提交
2562
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2563 2564
        super(InterpolationLayer, self).__init__(
            name, 'interpolation', 0, inputs=inputs, device=device)
Q
qijun 已提交
2565 2566
        config_assert(
            len(self.inputs) == 3, 'InterpolationLayer must have 3 inputs')
Z
zhangjinchao01 已提交
2567 2568 2569 2570 2571 2572 2573 2574
        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
2575

L
liaogang 已提交
2576 2577
@config_layer('bilinear_interp')
class BilinearInterpLayer(LayerBase):
Q
qijun 已提交
2578
    def __init__(self, name, inputs, **xargs):
L
liaogang 已提交
2579
        super(BilinearInterpLayer, self).__init__(
L
liaogang 已提交
2580
            name, 'bilinear_interp', 0, inputs=inputs, **xargs)
L
liaogang 已提交
2581
        input_layer = self.get_input_layer(0)
L
Luo Tao 已提交
2582 2583 2584 2585
        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
2586

L
liaogang 已提交
2587

Z
zhangjinchao01 已提交
2588 2589
@config_layer('sum_to_one_norm')
class SumToOneNormLayer(LayerBase):
Q
qijun 已提交
2590
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2591
        super(SumToOneNormLayer, self).__init__(
Q
qijun 已提交
2592 2593 2594
            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
2595 2596 2597
        input_layer0 = self.get_input_layer(0)
        self.set_layer_size(input_layer0.size)

Q
qijun 已提交
2598

Z
zhangjinchao01 已提交
2599 2600
@config_layer('cos_vm')
class CosSimVecMatLayer(LayerBase):
Q
qijun 已提交
2601
    def __init__(self, name, size, inputs, cos_scale=1.0, device=None):
Z
zhangjinchao01 已提交
2602
        super(CosSimVecMatLayer, self).__init__(
Q
qijun 已提交
2603
            name, 'cos_vm', size, inputs=inputs, device=device)
Z
zhangjinchao01 已提交
2604
        self.config.cos_scale = cos_scale
Q
qijun 已提交
2605 2606
        config_assert(
            len(self.inputs) == 2, 'CosSimVecMatLayer must have 2 inputs')
2607 2608 2609
        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
2610

Q
qijun 已提交
2611

Z
zhangjinchao01 已提交
2612 2613
@config_layer('sampling_id')
class SamplingIdLayer(LayerBase):
Q
qijun 已提交
2614
    def __init__(self, name, inputs, device=None):
Z
zhangjinchao01 已提交
2615 2616
        super(SamplingIdLayer, self).__init__(
            name, 'sampling_id', 0, inputs=inputs, device=device)
Q
qijun 已提交
2617 2618
        config_assert(
            len(self.inputs) == 1, 'SamplingIdLayer must have 1 input')
Z
zhangjinchao01 已提交
2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630
        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
2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
    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
2646
        self.config.average_strategy = average_strategy
Q
qijun 已提交
2647
        self.config.trans_type = trans_type
Z
zhangjinchao01 已提交
2648 2649 2650 2651 2652 2653
        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
2654

Z
zhangjinchao01 已提交
2655 2656
@config_layer('cos')
class CosSimLayer(LayerBase):
Q
qijun 已提交
2657
    def __init__(self, name, inputs, cos_scale=5, device=None):
Z
zhangjinchao01 已提交
2658 2659 2660 2661 2662 2663
        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')
2664
        self.config.cos_scale = cos_scale
Z
zhangjinchao01 已提交
2665 2666 2667 2668


@config_layer('tensor')
class TensorLayer(LayerBase):
Q
qijun 已提交
2669 2670 2671
    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
2672 2673
        config_assert(len(self.inputs) == 2, 'TensorLayer must have 2 inputs')
        config_assert(size > 0, 'size must be positive')
Q
qijun 已提交
2674 2675
        config_assert(inputs[1].parameter_name == None,
                      'second parameter should be None.')
Z
zhangjinchao01 已提交
2676 2677 2678 2679 2680 2681 2682 2683 2684 2685
        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
2686 2687 2688 2689 2690 2691 2692
    def __init__(self,
                 name,
                 inputs,
                 size=0,
                 bias=True,
                 error_clipping_threshold=None,
                 **xargs):
Z
zhangjinchao01 已提交
2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709
        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)
2710
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2711 2712 2713
                size = operator.calc_output_size(operator_conf.input_sizes)
                if size != 0:
                    self.set_layer_size(size)
2714
            else:
2715 2716
                sz = operator.calc_output_size(operator_conf.input_sizes)
                if sz != 0:
Q
qijun 已提交
2717 2718 2719 2720
                    config_assert(
                        sz == self.config.size,
                        "different inputs have different size: %s vs. %s" %
                        (sz, self.config.size))
Z
zhangjinchao01 已提交
2721 2722 2723 2724
        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
2725 2726 2727
                config_assert(
                    isinstance(input, Projection),
                    "input should be projection or operation")
2728
            if self.config.size == 0 and isinstance(input, Projection):
Z
zhangjinchao01 已提交
2729 2730 2731
                size = input.calc_output_size(input_layer)
                if size != 0:
                    self.set_layer_size(size)
2732
            elif isinstance(input, Projection):
Q
qijun 已提交
2733 2734 2735 2736 2737 2738
                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
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749
        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
2750 2751
                input_config.proj_conf.name = gen_parameter_name(name,
                                                                 input_index)
Z
zhangjinchao01 已提交
2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762
                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)

2763 2764 2765 2766 2767 2768
        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
2769

2770 2771 2772
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
Z
zhangjinchao01 已提交
2773

2774 2775
        if error_clipping_threshold is not None:
            self.config.error_clipping_threshold = error_clipping_threshold
Z
zhangjinchao01 已提交
2776

Q
qijun 已提交
2777

Z
zhangjinchao01 已提交
2778 2779
# like MixedLayer, but no bias parameter
@config_func
Q
qijun 已提交
2780
def ExpressionLayer(name, inputs, **xargs):
Z
zhangjinchao01 已提交
2781 2782
    MixedLayer(name, inputs, bias=False, **xargs)

Q
qijun 已提交
2783

Z
zhangjinchao01 已提交
2784 2785
@config_layer('concat')
class ConcatenateLayer(LayerBase):
Q
qijun 已提交
2786
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2787
        config_assert(inputs, 'inputs cannot be empty')
2788
        config_assert(not bias, 'ConcatenateLayer cannot support bias.')
Z
zhangjinchao01 已提交
2789 2790 2791 2792 2793 2794
        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
2795
            if self.config.size == 0:
Z
zhangjinchao01 已提交
2796 2797 2798 2799
                size += input_layer.size

        self.set_layer_size(size)

Q
qijun 已提交
2800

Z
zhangjinchao01 已提交
2801 2802 2803
# like concat layer, but each input layer was processed by a Projection.
@config_layer('concat2')
class ConcatenateLayer2(LayerBase):
Q
qijun 已提交
2804
    def __init__(self, name, inputs, bias=False, **xargs):
Z
zhangjinchao01 已提交
2805 2806 2807
        config_assert(inputs, 'inputs cannot be empty')
        super(ConcatenateLayer2, self).__init__(
            name, 'concat2', 0, inputs=inputs, **xargs)
2808 2809

        if isinstance(self.inputs[0], ConvProjection):
Q
qijun 已提交
2810 2811 2812 2813 2814 2815
            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.")
2816

Z
zhangjinchao01 已提交
2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836
        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
2837
                                              input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2838
            dims = input.calc_parameter_dims(input.proj_conf.input_size,
Q
qijun 已提交
2839
                                             input.proj_conf.output_size)
Z
zhangjinchao01 已提交
2840 2841
            self.create_input_parameter(input_index, psize, dims)

2842 2843 2844 2845 2846 2847 2848
        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()

2849 2850 2851
        if bias:
            self.config.bias_size = psize
            self.create_bias_parameter(bias, psize)
2852

Q
qijun 已提交
2853

Z
zhangjinchao01 已提交
2854 2855
@config_layer('recurrent')
class RecurrentLayer(LayerBase):
Q
qijun 已提交
2856
    def __init__(self, name, inputs, reversed=False, bias=True, **xargs):
Y
Yu Yang 已提交
2857 2858
        super(RecurrentLayer, self).__init__(name, 'recurrent', 0, inputs,
                                             **xargs)
Z
zhangjinchao01 已提交
2859 2860 2861 2862 2863 2864 2865 2866 2867
        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
2868

Z
zhangjinchao01 已提交
2869 2870
@config_layer('lstmemory')
class LstmLayer(LayerBase):
Q
qijun 已提交
2871 2872 2873 2874 2875 2876 2877 2878
    def __init__(self,
                 name,
                 inputs,
                 reversed=False,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Z
zhangjinchao01 已提交
2879 2880 2881 2882 2883 2884 2885 2886
        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
2887
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2888 2889 2890 2891 2892
        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
2893

Z
zhangjinchao01 已提交
2894 2895
@config_layer('lstm_step')
class LstmStepLayer(LayerBase):
Q
qijun 已提交
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905
    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
2906 2907 2908
        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
2909 2910 2911 2912 2913
        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
2914 2915 2916
        self.config.active_state_type = active_state_type
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2917

Z
zhangjinchao01 已提交
2918 2919 2920
# get the specific output from the input layer.
@config_layer('get_output')
class GetOutputLayer(LayerBase):
Q
qijun 已提交
2921 2922 2923 2924
    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
2925 2926 2927 2928
        inputs = self.inputs[0]
        config_assert(inputs.input_layer_argument,
                      'input_layer_argument cannot be empty')

Q
qijun 已提交
2929

Z
zhangjinchao01 已提交
2930 2931
@config_layer('mdlstmemory')
class MDLstmLayer(LayerBase):
Q
qijun 已提交
2932 2933 2934 2935 2936 2937 2938 2939
    def __init__(self,
                 name,
                 inputs,
                 directions=True,
                 active_gate_type="sigmoid",
                 active_state_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2940 2941
        super(MDLstmLayer, self).__init__(name, 'mdlstmemory', 0, inputs,
                                          **xargs)
Z
zhangjinchao01 已提交
2942 2943 2944 2945
        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
2946 2947
        config_assert(input_layer.size % (3 + dim_num) == 0,
                      "size % (dim_num) should be 0!")
Q
qijun 已提交
2948
        size = input_layer.size / (3 + dim_num)
Z
zhangjinchao01 已提交
2949
        self.set_layer_size(size)
Q
qijun 已提交
2950
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2951 2952 2953
        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
2954 2955
        self.create_input_parameter(0, size * size * (3 + dim_num),
                                    [size, size, 3 + dim_num])
Z
zhangjinchao01 已提交
2956
        #bias includes 3 kinds of peephole, 3+dim_num+2+dim_num
Q
qijun 已提交
2957 2958
        self.create_bias_parameter(bias, size * (5 + 2 * dim_num))

Z
zhangjinchao01 已提交
2959 2960 2961

@config_layer('gated_recurrent')
class GatedRecurrentLayer(LayerBase):
Q
qijun 已提交
2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972
    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
2973 2974 2975 2976 2977 2978
        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
2979
        self.config.active_gate_type = active_gate_type
Z
zhangjinchao01 已提交
2980 2981 2982
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
2983

Z
zhangjinchao01 已提交
2984 2985
@config_layer('gru_step')
class GruStepLayer(LayerBase):
Q
qijun 已提交
2986 2987 2988 2989 2990 2991 2992
    def __init__(self,
                 name,
                 size,
                 inputs,
                 active_gate_type="sigmoid",
                 bias=True,
                 **xargs):
Y
Yu Yang 已提交
2993 2994
        super(GruStepLayer, self).__init__(name, 'gru_step', size, inputs,
                                           **xargs)
Z
zhangjinchao01 已提交
2995 2996 2997
        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
2998 2999 3000 3001 3002
        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
3003 3004 3005
        self.create_input_parameter(0, size * size * 3, [size, size * 3])
        self.create_bias_parameter(bias, size * 3)

Q
qijun 已提交
3006

Z
zhangjinchao01 已提交
3007 3008 3009 3010 3011 3012 3013
'''
 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
3014 3015


Z
zhangjinchao01 已提交
3016 3017
@config_layer('crf')
class CRFLayer(LayerBase):
Q
qijun 已提交
3018
    def __init__(self, name, size, inputs, coeff=1.0, device=None):
Z
zhangjinchao01 已提交
3019
        super(CRFLayer, self).__init__(name, 'crf', size, inputs, device=device)
Q
qijun 已提交
3020 3021
        config_assert(2 <= len(self.inputs) <= 3,
                      'CRFLayer must have 2 or 3 inputs')
Z
zhangjinchao01 已提交
3022 3023 3024
        self.create_input_parameter(0, size * (size + 2), [size, size + 2])
        self.config.coeff = coeff

Q
qijun 已提交
3025

Z
zhangjinchao01 已提交
3026 3027 3028 3029 3030 3031 3032 3033
'''
 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
3034 3035


Z
zhangjinchao01 已提交
3036 3037
@config_layer('crf_decoding')
class CRFDecodingLayer(LayerBase):
Q
qijun 已提交
3038
    def __init__(self, name, size, inputs, device=None):
Z
zhangjinchao01 已提交
3039 3040 3041 3042 3043 3044 3045
        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
3046

Z
zhangjinchao01 已提交
3047 3048
@config_layer('ctc')
class CTCLayer(LayerBase):
Q
qijun 已提交
3049
    def __init__(self, name, size, inputs, norm_by_times=False, device=None):
Z
zhangjinchao01 已提交
3050 3051 3052 3053
        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
3054

3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
@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
3076 3077
@config_layer('recurrent_layer_group')
class RecurrentLayerGroup(LayerBase):
Q
qijun 已提交
3078
    def __init__(self, name, device=None):
L
Luo Tao 已提交
3079 3080
        global g_pass_height_width
        g_pass_height_width = False
Z
zhangjinchao01 已提交
3081 3082 3083 3084 3085 3086
        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
3087
def Layer(name, type, **xargs):
Z
zhangjinchao01 已提交
3088 3089 3090 3091
    layers = {}
    layers.update(g_cost_map)
    layers.update(g_layer_type_map)
    layer_func = layers.get(type)
Q
qijun 已提交
3092
    config_assert(layer_func, "layer type '%s' not supported." % type)
X
xuwei06 已提交
3093
    return layer_func(name, **xargs)
Z
zhangjinchao01 已提交
3094

Q
qijun 已提交
3095

Z
zhangjinchao01 已提交
3096
@config_func
Q
qijun 已提交
3097
def ParameterHook(type, **kwargs):
Z
zhangjinchao01 已提交
3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
    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
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131
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
3132 3133 3134 3135 3136 3137 3138

    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
3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149
    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
3150 3151
    config_assert(not momentum or not decay_rate_l1,
                  "momentum and decay_rate_l1 cannot both be non-zero")
3152 3153 3154 3155 3156

    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
3157 3158 3159 3160
    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)
3161

Q
qijun 已提交
3162 3163
    num_batches_regularization = default(num_batches_regularization,
                                         g_default_num_batches_regularization)
3164 3165 3166
    if num_batches_regularization is not None:
        para.num_batches_regularization = int(num_batches_regularization)

Z
zhangjinchao01 已提交
3167 3168 3169 3170 3171 3172
    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
3173 3174
    gradient_clipping_threshold = default(gradient_clipping_threshold,
                                          g_default_gradient_clipping_threshold)
3175 3176
    if gradient_clipping_threshold is not None:
        para.gradient_clipping_threshold = gradient_clipping_threshold
Q
qijun 已提交
3177 3178
    para.initial_strategy = default(initial_strategy,
                                    g_default_initial_strategy)
Z
zhangjinchao01 已提交
3179 3180 3181 3182 3183 3184
    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
3185 3186 3187
            print(
                "Use initial_smart, but dims not set. Initial_smart may not be used in this layer"
            )
Z
zhangjinchao01 已提交
3188 3189 3190 3191
            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)
3192 3193 3194 3195 3196 3197 3198

    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
3199 3200 3201 3202
    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")
3203 3204
    if is_shared is not None:
        para.is_shared = is_shared
Z
zhangjinchao01 已提交
3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225

    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
3226

Z
zhangjinchao01 已提交
3227 3228 3229 3230 3231
@config_func
def default_initial_mean(val):
    global g_default_initial_mean
    g_default_initial_mean = val

Q
qijun 已提交
3232

Z
zhangjinchao01 已提交
3233 3234 3235 3236 3237
@config_func
def default_initial_strategy(val):
    global g_default_initial_strategy
    g_default_initial_strategy = val

Q
qijun 已提交
3238

Z
zhangjinchao01 已提交
3239 3240 3241 3242 3243
@config_func
def default_initial_smart(val):
    global g_default_initial_smart
    g_default_initial_smart = val

Q
qijun 已提交
3244

Z
zhangjinchao01 已提交
3245 3246 3247 3248 3249
@config_func
def default_momentum(val):
    global g_default_momentum
    g_default_momentum = val

Q
qijun 已提交
3250

Z
zhangjinchao01 已提交
3251 3252 3253 3254 3255
@config_func
def default_decay_rate(val):
    global g_default_decay_rate
    g_default_decay_rate = val

Q
qijun 已提交
3256

Z
zhangjinchao01 已提交
3257 3258 3259 3260 3261
@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
3262

Z
zhangjinchao01 已提交
3263 3264 3265 3266 3267
@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
3268

Z
zhangjinchao01 已提交
3269 3270 3271 3272 3273
@config_func
def default_device(val):
    global g_default_device
    g_default_device = val

Q
qijun 已提交
3274

Z
zhangjinchao01 已提交
3275 3276 3277 3278 3279
@config_func
def default_update_hooks(val):
    global g_default_update_hooks
    g_default_update_hooks = val

Q
qijun 已提交
3280

Z
zhangjinchao01 已提交
3281 3282 3283 3284 3285
@config_func
def default_compact_func(val):
    global g_default_compact_func
    g_default_compact_func = val

Q
qijun 已提交
3286

Z
zhangjinchao01 已提交
3287 3288 3289 3290 3291
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
3292 3293 3294
        execfile(config_file,
                 make_config_environment(config_file, config_args), local_args)

Z
zhangjinchao01 已提交
3295 3296
    return Import

Q
qijun 已提交
3297

Z
zhangjinchao01 已提交
3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
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
3326 3327 3328
    adam_beta1=0.9,
    adam_beta2=0.999,
    adam_epsilon=1e-8, )
Z
zhangjinchao01 已提交
3329

Q
qijun 已提交
3330
settings_deprecated = dict(usage_ratio=1., )
Z
zhangjinchao01 已提交
3331 3332 3333 3334

trainer_settings = dict(
    save_dir="./output/model",
    init_model_path=None,
Q
qijun 已提交
3335 3336
    start_pass=0, )

Z
zhangjinchao01 已提交
3337 3338 3339 3340 3341

@config_func
def Settings(**args):
    for k, v in args.iteritems():
        if k == "usage_ratio":
Q
qijun 已提交
3342 3343
            logger.warning(
                "Deprecated: define usage_ratio in DataConfig instead")
Z
zhangjinchao01 已提交
3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354
            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
3355

Z
zhangjinchao01 已提交
3356 3357 3358 3359
@config_func
def cluster_config(**args):
    pass

Q
qijun 已提交
3360

Z
zhangjinchao01 已提交
3361 3362 3363 3364 3365 3366 3367 3368 3369
@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
3370

Z
zhangjinchao01 已提交
3371 3372 3373 3374
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
3375

Z
zhangjinchao01 已提交
3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390
        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
3391
        get_config_arg=make_get_config_arg(config_args), )
Z
zhangjinchao01 已提交
3392 3393 3394 3395 3396

    funcs.update(g_extended_config_funcs)

    return funcs

Q
qijun 已提交
3397

Z
zhangjinchao01 已提交
3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413
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
3414

Z
zhangjinchao01 已提交
3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426
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
3427

Z
zhangjinchao01 已提交
3428 3429 3430 3431
def my_fatal(s):
    logger.critical(s)
    raise Exception()

Y
Yu Yang 已提交
3432

3433
_parse_config_hooks = set()
Y
Yu Yang 已提交
3434 3435


3436 3437 3438 3439 3440 3441 3442
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
3443

Y
Yu Yang 已提交
3444

3445
def update_g_config():
Z
zhangjinchao01 已提交
3446
    '''
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 3473
    Update g_config after execute config_file or config_functions.
    '''
    for k, v in settings.iteritems():
        if v is None:
            continue
        g_config.opt_config.__setattr__(k, v)

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

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


def parse_config(trainer_config, config_arg_str):
    '''
    @param trainer_config: can be a string of config file name or a function name
    with config logic
Z
zhangjinchao01 已提交
3474 3475 3476 3477
    @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()
3478 3479
    for hook in _parse_config_hooks:
        hook()
Z
zhangjinchao01 已提交
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506

    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

3507 3508
    if hasattr(trainer_config, '__call__'):
        trainer_config.func_globals.update(
L
Luo Tao 已提交
3509
            make_config_environment("", config_args))
3510
        trainer_config()
H
hanchao 已提交
3511
    else:
3512 3513
        execfile(trainer_config,
                 make_config_environment(trainer_config, config_args))
Z
zhangjinchao01 已提交
3514

3515
    return update_g_config()
Z
zhangjinchao01 已提交
3516 3517


3518
def parse_config_and_serialize(trainer_config, config_arg_str):
Z
zhangjinchao01 已提交
3519
    try:
3520
        config = parse_config(trainer_config, config_arg_str)
Z
zhangjinchao01 已提交
3521 3522 3523 3524 3525 3526
        #logger.info(config)
        return config.SerializeToString()
    except:
        traceback.print_exc()
        raise

Q
qijun 已提交
3527

Z
zhangjinchao01 已提交
3528 3529 3530 3531 3532 3533 3534 3535
if __name__ == '__main__':
    try:
        config = parse_config(sys.argv[1], '')
        config.SerializeToString()
        __real_print__(str(config))
    except:
        traceback.print_exc()
        raise