backward.py 32.7 KB
Newer Older
1
#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
D
dzhwinter 已提交
2
#
D
dzhwinter 已提交
3 4 5
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
D
dzhwinter 已提交
6
#
D
dzhwinter 已提交
7
#     http://www.apache.org/licenses/LICENSE-2.0
D
dzhwinter 已提交
8
#
D
dzhwinter 已提交
9 10 11 12 13 14
# 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.

15 16
from __future__ import print_function

17
from paddle.fluid import framework as framework
F
update  
fengjiayi 已提交
18
from . import core
F
update  
fengjiayi 已提交
19
import collections
20
import copy
21
import six
M
minqiyang 已提交
22
from .. import compat as cpt
23
from . import unique_name
24

25
__all__ = ['append_backward', 'gradients']
26 27


28 29
def _rename_arg_(op_descs, old_name, new_name, begin_idx=None, end_idx=None):
    """
30
    Traverse all ops in op_descs[begin_idx : end_idx],
31 32
    if any op has inputs/outputs named "old_name", rename it as 'new_name'
    """
F
update  
fengjiayi 已提交
33 34 35
    if begin_idx is None:
        begin_idx = 0
    if end_idx is None:
36
        end_idx = len(op_descs)
F
update  
fengjiayi 已提交
37
    for i in range(begin_idx, end_idx):
38
        op_desc = op_descs[i]
F
fengjiayi 已提交
39 40
        if isinstance(op_desc, tuple):
            op_desc = op_desc[0]
W
Wu Yi 已提交
41 42
        op_desc._rename_input(old_name, new_name)
        op_desc._rename_output(old_name, new_name)
F
update  
fengjiayi 已提交
43 44


F
fengjiayi 已提交
45
def _create_op_desc_(op_type, inputs, outputs, attrs):
46 47 48
    """
    Create a C++ OpDesc object with specified inputs, outputs and attributes.
    """
F
fengjiayi 已提交
49 50
    op_desc = core.OpDesc()
    op_desc.set_type(op_type)
M
minqiyang 已提交
51
    for para, args in six.iteritems(inputs):
52 53 54 55 56
        op_desc.set_input(
            para,
            list(
                map(lambda arg: arg.decode() if isinstance(arg, six.binary_type) else arg,
                    args)))
M
minqiyang 已提交
57
    for para, args in six.iteritems(outputs):
58 59 60 61 62
        op_desc.set_output(
            para,
            list(
                map(lambda arg: arg.decode() if isinstance(arg, six.binary_type) else arg,
                    args)))
Y
yuyang18 已提交
63 64 65 66 67 68

    op_role_attr_name = core.op_proto_and_checker_maker.kOpRoleAttrName()

    if op_role_attr_name not in attrs:
        attrs[
            op_role_attr_name] = core.op_proto_and_checker_maker.OpRole.Backward
M
minqiyang 已提交
69
    for name, val in six.iteritems(attrs):
F
fengjiayi 已提交
70 71 72
        if isinstance(val, framework.Block):
            op_desc.set_block_attr(name, val.desc)
        else:
W
Wu Yi 已提交
73
            op_desc._set_attr(name, val)
F
fengjiayi 已提交
74 75 76
    return op_desc


77 78 79 80
def _infer_var_data_type_(grad_var_name, block):
    """
    Infer the data type of given grad variable
    """
M
minqiyang 已提交
81 82 83 84
    grad_var = block.desc.find_var(cpt.to_bytes(grad_var_name))
    fwd_name = _strip_grad_suffix_(grad_var_name)
    if block.desc.has_var_recursive(cpt.to_bytes(fwd_name)):
        fwd_var = block.desc.find_var_recursive(cpt.to_bytes(fwd_name))
F
fengjiayi 已提交
85 86
        grad_var.set_dtype(fwd_var.dtype())
    else:
87
        grad_var.set_dtype(core.VarDesc.VarType.FP32)
F
fengjiayi 已提交
88 89


F
fengjiayi 已提交
90
def _all_in_set_(cands, s):
91 92 93
    """
    Test if all elements of 'cands' are in set 's'
    """
F
fengjiayi 已提交
94 95
    if len(cands) == 0:
        return False
F
fengjiayi 已提交
96 97 98 99 100 101
    for c in cands:
        if not c in s:
            return False
    return True


102 103 104 105 106 107
def _some_in_set_(cands, s):
    """
    Test if some elements of 'cands' are in set 's'
    """
    if len(cands) == 0:
        return False
M
minqiyang 已提交
108 109
    literal_set = cpt.to_text(s)
    literal_cands = cpt.to_text(cands)
M
minqiyang 已提交
110 111
    for c in literal_cands:
        if c in literal_set:
112 113 114 115
            return True
    return False


F
fengjiayi 已提交
116
def _strip_grad_suffix_(name):
117 118 119 120 121
    """
    Strip the grad suffix from the given varibale name
    e.g. x@GRAD ==> x
         y@GRAD@RENAME@1 ==> y
    """
M
minqiyang 已提交
122
    name = cpt.to_text(name)
M
minqiyang 已提交
123
    pos = name.find(core.grad_var_suffix())
F
fengjiayi 已提交
124
    return name[:pos] if pos != -1 else name
F
fengjiayi 已提交
125 126 127


def _append_grad_suffix_(name):
128 129 130 131
    """
    Append grad suffix to the given variable name
    e.g. x ==> x@GRAD
    """
M
minqiyang 已提交
132
    return cpt.to_text(name) + core.grad_var_suffix()
F
fengjiayi 已提交
133 134


F
fengjiayi 已提交
135
def _addup_repetitive_outputs_(op_descs):
136 137
    """
    In backward part, an variable may be the output of more than one ops.
F
fengjiayi 已提交
138 139
    And one op may yield its multiple outputs to the same variable.
    In these cases, the variable should be the accumulation of all the outputs.
140 141
    `sum_op`s are added to implement the accumulate.
    """
F
update  
fengjiayi 已提交
142 143
    pending_sum_ops = []
    var_rename_count = collections.defaultdict(int)
F
fengjiayi 已提交
144
    renamed_vars = collections.defaultdict(list)
145
    renamed_var_start_idx = collections.defaultdict(list)
F
fengjiayi 已提交
146
    for idx, op_desc in enumerate(op_descs):
F
update  
fengjiayi 已提交
147
        for var_name in op_desc.input_arg_names():
F
fengjiayi 已提交
148
            if len(renamed_vars[var_name]) > 1:
149 150 151
                pending_sum_ops.append((_create_op_desc_(
                    "sum", {"X": renamed_vars[var_name]}, {"Out": [var_name]},
                    {"use_mkldnn": False}), idx))
F
fengjiayi 已提交
152
                renamed_vars[var_name] = [var_name]
F
update  
fengjiayi 已提交
153
        for param_idx, param_name in enumerate(op_desc.output_names()):
F
fengjiayi 已提交
154 155 156 157 158 159 160 161 162
            arg_names = op_desc.output(param_name)
            for arg_idx, var_name in enumerate(arg_names):
                if var_name == core.empty_var_name(
                ) or var_name in op_desc.input_arg_names():
                    # empty variable or inplace op
                    continue
                if len(renamed_vars[var_name]) == 0:
                    # it's the first time we get the variable
                    renamed_vars[var_name] = [var_name]
163
                    renamed_var_start_idx[var_name] = idx
F
fengjiayi 已提交
164 165 166 167 168 169 170
                else:
                    if len(renamed_vars[var_name]) == 1:
                        new_name = var_name + "@RENAME@" + \
                            str(var_rename_count[var_name])
                        var_rename_count[var_name] += 1
                        # rename original var_name
                        renamed_vars[var_name][0] = new_name
171 172 173 174 175 176
                        # before change: _rename_arg_(op_descs, var_name,
                        #                             new_name, 0, idx)
                        # rename arg from idx of the first appearance
                        # in backward, not always from 0
                        _rename_arg_(op_descs, var_name, new_name,
                                     renamed_var_start_idx[var_name], idx)
F
fengjiayi 已提交
177 178
                        _rename_arg_(pending_sum_ops, var_name, new_name)

F
update  
fengjiayi 已提交
179 180 181 182 183 184 185 186 187 188 189 190 191
                        for p in op_desc.output_names()[:param_idx]:
                            p_arg_names = op_desc.output(p)
                            if var_name in p_arg_names:
                                op_desc.set_output(p, [
                                    new_name if x == var_name else x
                                    for x in p_arg_names
                                ])

                        arg_names = [
                            new_name if x == var_name else x
                            for x in arg_names[:arg_idx]
                        ] + arg_names[arg_idx:]

F
update  
fengjiayi 已提交
192
                    new_name = var_name + "@RENAME@" + \
F
fengjiayi 已提交
193
                        str(var_rename_count[var_name])
F
fengjiayi 已提交
194
                    var_rename_count[var_name] += 1
F
fengjiayi 已提交
195 196 197
                    arg_names[arg_idx] = new_name
                    op_desc.set_output(param_name, arg_names)
                    renamed_vars[var_name].append(new_name)
F
update  
fengjiayi 已提交
198

M
minqiyang 已提交
199
    for var_name, inputs in six.iteritems(renamed_vars):
F
update  
fengjiayi 已提交
200
        if len(inputs) > 1:
201 202 203
            pending_sum_ops.append(
                (_create_op_desc_("sum", {"X": inputs}, {"Out": [var_name]},
                                  {"use_mkldnn": False}), len(op_descs)))
F
fengjiayi 已提交
204
    # sum_op descs are sorted according to their insert position
F
update  
fengjiayi 已提交
205
    for p in reversed(pending_sum_ops):
F
fengjiayi 已提交
206 207 208 209 210 211
        op_descs.insert(p[1], p[0])

    return op_descs


def _remove_no_grad_branch_(op_descs, no_grad_set):
212 213 214 215
    """
    Remove unnecessary grad ops
    A grad op can be removed in two cases:
        1. all outputs of the grad op are in 'no_grad_set'
F
fengjiayi 已提交
216
        2. all grad inputs of the grad op are in 'no_grad_set'
217
    """
F
fengjiayi 已提交
218 219

    def _op_can_be_removed_(op_desc, no_grad_set):
F
fengjiayi 已提交
220 221
        out_arg_names = op_desc.output_arg_names()
        if len(out_arg_names) == 0 or _all_in_set_(out_arg_names, no_grad_set):
F
fengjiayi 已提交
222
            return True
223 224 225 226
        if _all_in_set_([
                name for name in op_desc.input_arg_names()
                if name.find(core.grad_var_suffix()) != -1
        ], no_grad_set):
F
fengjiayi 已提交
227
            no_grad_set.update(out_arg_names)
F
fengjiayi 已提交
228 229 230
            return True
        return False

F
fengjiayi 已提交
231
    # Remove ops whose outputs are all in no_grad_dict
232 233 234 235
    op_descs = [
        op_desc for op_desc in op_descs
        if not _op_can_be_removed_(op_desc, no_grad_set)
    ]
F
fengjiayi 已提交
236 237
    # Insert fill_zeros_like_op
    to_insert = []
F
fengjiayi 已提交
238
    for idx, op_desc in enumerate(op_descs):
F
fengjiayi 已提交
239
        for arg in op_desc.input_arg_names():
F
fengjiayi 已提交
240
            if core.grad_var_suffix() in arg and arg in no_grad_set:
241
                x_in = _strip_grad_suffix_(arg)
242 243
                to_insert.append((_create_op_desc_(
                    "fill_zeros_like", {"X": [x_in]}, {"Out": [arg]}, {}), idx))
F
fengjiayi 已提交
244

245
    list([op_descs.insert(p[1], p[0]) for p in reversed(to_insert)])
F
fengjiayi 已提交
246 247 248 249

    return op_descs


250
from .proto import framework_pb2
Y
Yang Yang 已提交
251 252 253 254


def serialize_op_decs(op_desc):
    protostr = op_desc.serialize_to_string()
M
minqiyang 已提交
255
    proto = framework_pb2.OpDesc.FromString(six.binary_type(protostr))
Y
Yang Yang 已提交
256 257 258
    return proto.__str__()


259 260
def _append_backward_ops_(block,
                          ops,
F
fengjiayi 已提交
261 262 263
                          target_block,
                          no_grad_dict,
                          grad_to_var,
264 265
                          callbacks=None,
                          input_grad_names_set=None):
266 267 268 269 270
    """
    Create all grad ops, and insert them into given block

    Args:
        block(Block): the block where forward ops are
271
        ops(Op): the forward operators whose backward ops need to be added
272
        target_block(Block): the block which is going to hold new generated grad ops
273
        no_grad_dict(dict):
274 275 276 277 278
            key(int)  block index
            val(set) a set of varibale names. These varibales have no gradient
        grad_to_var(dict)(output argument):
            key(str): grad variable name
            val(str): corresponding forward variable name
F
fengjiayi 已提交
279
        callback(callable object): a callable object used to decorate new generated grad ops
280
    """
Y
Yang Yang 已提交
281
    if callbacks is not None:
Y
Yang Yang 已提交
282 283 284 285
        assert (isinstance(callbacks, list))
        for cb in callbacks:
            if not hasattr(cb, '__call__'):
                raise ValueError("'callback' must be a callable object.")
F
fengjiayi 已提交
286

F
fengjiayi 已提交
287
    # grad_op_descs holds created grad_op, and will be appended to target_block
F
fengjiayi 已提交
288 289
    grad_op_descs = []
    program = block.program
290
    for op in reversed(ops):
F
fengjiayi 已提交
291 292 293
        grad_sub_block_list = []
        # If the op has its own sub-block, deal with the sub-block first
        if op.has_attr("sub_block"):
W
Wu Yi 已提交
294
            sub_block = program.block(op._block_attr_id("sub_block"))
W
Wu Yi 已提交
295
            grad_sub_block = program._create_block()
W
Wu Yi 已提交
296
            grad_sub_block._set_forward_block_idx(sub_block.idx)
297 298 299
            # see follwing comments for why set None here.
            pre_input_grad_names_set = copy.copy(input_grad_names_set)
            input_grad_names_set = None
X
Xin Pan 已提交
300
            _append_backward_ops_(sub_block, sub_block.ops, grad_sub_block,
301 302 303
                                  no_grad_dict, grad_to_var, callbacks,
                                  input_grad_names_set)
            input_grad_names_set = pre_input_grad_names_set
Y
Yu Yang 已提交
304

W
Wu Yi 已提交
305
            program._rollback()
F
fengjiayi 已提交
306 307
            grad_sub_block_list.append(grad_sub_block.desc)

F
fengjiayi 已提交
308
        # Getting op's corresponding grad_op
F
fengjiayi 已提交
309
        grad_op_desc, op_grad_to_var = core.get_grad_op_desc(
M
minqiyang 已提交
310
            op.desc, cpt.to_text(no_grad_dict[block.idx]), grad_sub_block_list)
Y
Yang Yu 已提交
311

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338
        # If input_grad_names_set is not None, extend grad_op_descs only when
        # any input grad in outputs of previous grad ops.
        # But this strategy is not suited for while op for some control flow,
        # for example, for while op, the grads maybe generated in next loop.
        if input_grad_names_set is not None:
            is_append_grad = False
            for op_desc in grad_op_desc:
                input_grad_names = [
                    name for name in op_desc.input_arg_names()
                    if name.find(core.grad_var_suffix()) != -1
                ]
                # some code of gradient ops, like increment, are not very
                # standard, there is no @GRAD in these ops' inputs.
                if len(input_grad_names) == 0:
                    is_append_grad = True
                    break

                if _some_in_set_(input_grad_names, input_grad_names_set):
                    grad_op_descs.append(op_desc)
                    is_append_grad = True
                    for name in op_desc.output_arg_names():
                        input_grad_names_set.add(name)
            if is_append_grad:
                grad_to_var.update(op_grad_to_var)
        else:
            grad_op_descs.extend(grad_op_desc)
            grad_to_var.update(op_grad_to_var)
F
fengjiayi 已提交
339 340 341 342 343

    grad_op_descs = _addup_repetitive_outputs_(grad_op_descs)

    grad_op_descs = _remove_no_grad_branch_(grad_op_descs,
                                            no_grad_dict[block.idx])
F
fengjiayi 已提交
344

F
fengjiayi 已提交
345
    # append op_desc in grad_op_descs to target_block
Y
yuyang18 已提交
346 347
    op_role_attr_name = core.op_proto_and_checker_maker.kOpRoleAttrName()
    backward = core.op_proto_and_checker_maker.OpRole.Backward
F
update  
fengjiayi 已提交
348
    for op_desc in grad_op_descs:
F
fengjiayi 已提交
349 350
        new_op_desc = target_block.desc.append_op()
        new_op_desc.copy_from(op_desc)
W
Wu Yi 已提交
351
        new_op_desc._set_attr(op_role_attr_name, backward)
Y
Yang Yang 已提交
352
        grad_to_var["__current_op_desc__"] = new_op_desc
Y
Yang Yang 已提交
353 354 355 356
        if callbacks is not None:
            assert (isinstance(callbacks, list))
            for cb in callbacks:
                cb(block=target_block, context=grad_to_var)
F
update  
fengjiayi 已提交
357

F
fengjiayi 已提交
358 359

def _append_backward_vars_(block, start_op_idx, grad_to_var, grad_info_map):
360 361 362 363 364 365 366 367 368 369 370 371
    """
    Create new variables required by backward pass.

    Args:
        block(Block): the block where new variables will be created
        start_op_idx(int): Only variables required by ops in block.ops[start_op_idx : ] will be created
        grad_to_var(dict):
            key(str): grad variable name
            val(str): corresponding forward variable name
            In most cases, this dict is generated by _append_backward_ops_()
        grad_info_map(dict)(output argument):
            key(str): forward variable name
372
            val(tuple): a tuple of (str, Block), str is the corresponding grad name, Block is the block containing grad variable
373
    """
F
fengjiayi 已提交
374 375 376
    for op_idx in range(start_op_idx, block.desc.op_size()):
        op_desc = block.desc.op(op_idx)
        if op_desc.has_attr("sub_block"):
W
Wu Yi 已提交
377
            sub_block = block.program.block(op_desc._block_attr_id("sub_block"))
F
fengjiayi 已提交
378 379 380 381
            _append_backward_vars_(sub_block, 0, grad_to_var, grad_info_map)
        new_vars = set()
        # create new gradient variables
        for grad_var_name in op_desc.output_arg_names():
M
minqiyang 已提交
382 383
            if block.desc.has_var_recursive(cpt.to_bytes(
                    grad_var_name)) or grad_var_name == core.empty_var_name():
F
fengjiayi 已提交
384
                continue
M
minqiyang 已提交
385
            block.desc.var(cpt.to_bytes(grad_var_name))
F
fengjiayi 已提交
386
            new_vars.add(grad_var_name)
387
            if grad_var_name not in grad_to_var:
F
fengjiayi 已提交
388 389 390 391 392 393 394 395
                continue
            grad_info_map[grad_to_var[grad_var_name]] = (grad_var_name, block)
        # infer_shape and infer_type
        op_desc.infer_var_type(block.desc)
        op_desc.infer_shape(block.desc)
        for arg in op_desc.output_arg_names():
            if arg in new_vars:
                _infer_var_data_type_(arg, block)
F
update  
fengjiayi 已提交
396 397


398 399 400 401 402 403
def _rename_grad_(block, start_op_idx, grad_to_var, target_grad_map):
    var_map = copy.copy(target_grad_map)
    for op_idx in range(start_op_idx, block.desc.op_size()):
        op_desc = block.desc.op(op_idx)
        for name in op_desc.input_arg_names():
            if name in var_map:
W
Wu Yi 已提交
404
                op_desc._rename_input(name, var_map[name])
405 406 407

        for name in op_desc.output_arg_names():
            if block.desc.find_var(name.encode("ascii")):
Y
Yu Yang 已提交
408
                new_name = unique_name.generate(name)
W
Wu Yi 已提交
409
                op_desc._rename_output(name, new_name)
410 411
                var_map[name] = new_name

M
minqiyang 已提交
412
    for g, ng in six.iteritems(var_map):
413 414 415 416 417 418 419 420 421 422 423
        if g in grad_to_var:
            grad_to_var[ng] = grad_to_var[g]
            grad_to_var.pop(g)


def _get_stop_gradients_(program):
    no_grad_dict = dict()
    assert isinstance(program, framework.Program)
    for block in program.blocks:
        assert isinstance(block, framework.Block)
        block_no_grad_set = set()
424
        for var in list(block.vars.values()):
425 426 427 428 429 430 431
            assert isinstance(var, framework.Variable)
            if var.stop_gradient:
                block_no_grad_set.add(_append_grad_suffix_(var.name))
        no_grad_dict[block.idx] = block_no_grad_set
    return no_grad_dict


Y
Yang Yang 已提交
432 433
def append_backward(loss, parameter_list=None, no_grad_set=None,
                    callbacks=None):
434
    """
F
fengjiayi 已提交
435 436
    Append backward part to main_program.

437 438 439
    A complete neural network training is made up of forward and backward
    propagation. However, when we configure a network, we only need to
    specify its forwrd part. The backward part is generated automatically
F
fengjiayi 已提交
440 441
    according to the forward part by this function.

442
    In most cases, users do not need to invoke this function manually. It
F
fengjiayi 已提交
443
    will be automatically invoked by the optimizer's `minimize` function.
F
fengjiayi 已提交
444 445

    Args:
F
fengjiayi 已提交
446
        loss(Variable): The loss variable of the network.
447 448 449
        parameter_list(list[string]|None): Names of parameters that need
                                           to be updated by optimizers.
                                           If it is None, all parameters
F
fengjiayi 已提交
450 451
                                           will be updated.
                                           Default: None
452 453 454
        no_grad_set(set|None): Variables in the Block 0 whose gradients
                               should be ignored. All variables with
                               `step_gradient=True` from all blocks will
F
fengjiayi 已提交
455 456
                               be automatically added into this set.
                               Default: None
457 458 459 460 461 462 463 464 465 466 467 468 469 470
        callbacks(list[callable object]|None): The callbacks are used for
                                               doing some custom jobs during
                                               backward part building. All
                                               callable objects in it will
                                               be invoked once each time a
                                               new gradient operator is added
                                               into the program. The callable
                                               object must has two input
                                               parameters: 'block' and 'context'.
                                               The 'block' is the block which
                                               the new gradient operator will
                                               be added to. The 'context' is a
                                               map, whose keys are gradient
                                               variable names and values are
F
fengjiayi 已提交
471
                                               corresponding original variables.
472 473 474 475 476 477
                                               In addition to this, the 'context'
                                               has another special key-value pair:
                                               the key is string '__current_op_desc__'
                                               and the value is the op_desc of the
                                               gradient operator who has just
                                               triggered the callable object.
F
fengjiayi 已提交
478 479

    Returns:
480 481
        list[(Variable,Variable)]: Pairs of parameter and its
        corresponding gradients. The key is the parameter and the
F
fengjiayi 已提交
482 483 484 485 486 487 488 489
        value is gradient variable.

    Raises:
        AssertionError: If `loss` is not an instance of Variable.

    Examples:
        .. code-block:: python

F
fengjiayi 已提交
490
            # network configuration code
L
lujun 已提交
491
            # loss from ...
492
            import paddle.fluid as fluid
L
lujun 已提交
493 494 495 496 497 498
            x = fluid.layers.data(name='x', shape=[13], dtype='float32')
            y = fluid.layers.data(name='y', shape=[1], dtype='float32')

            y_predict = fluid.layers.fc(input=x, size=1, act=None)
            loss = fluid.layers.square_error_cost(input=y_predict, label=y)

F
fengjiayi 已提交
499 500
            avg_loss = fluid.layers.mean(loss)
            param_grad_list = fluid.backward.append_backward(loss=avg_loss)
501 502
    """
    assert isinstance(loss, framework.Variable)
Y
yuyang18 已提交
503

Y
Fix bug  
yuyang18 已提交
504 505 506 507 508 509 510 511 512 513 514
    if loss.op is None:
        # the loss is from a cloned program. Find loss op manually.
        for op in reversed(loss.block.ops):
            assert isinstance(op, framework.Operator)
            if len(op.output_arg_names) == 1 and op.output_arg_names[
                    0] == loss.name:
                loss.op = op
                break
        if loss.op is None:
            raise ValueError("loss.op is None. Should not happend")

W
Wu Yi 已提交
515 516 517
    loss.op._set_attr(core.op_proto_and_checker_maker.kOpRoleAttrName(),
                      int(core.op_proto_and_checker_maker.OpRole.Forward) |
                      int(core.op_proto_and_checker_maker.OpRole.Loss))
Y
yuyang18 已提交
518

Y
Yang Yang 已提交
519 520
    if callbacks is not None:
        isinstance(callbacks, list)
Y
Yu Yang 已提交
521

F
fengjiayi 已提交
522
    program = loss.block.program
523 524
    program._appending_grad_times += 1

F
fengjiayi 已提交
525
    if no_grad_set is None:
526 527 528
        no_grad_set = set()
    no_grad_set = copy.copy(no_grad_set)
    no_grad_dict = _get_stop_gradients_(program)
529
    no_grad_dict[0].update(list(map(_append_grad_suffix_, no_grad_set)))
Y
Yu Yang 已提交
530

F
update  
fengjiayi 已提交
531
    grad_info_map = dict()
F
fengjiayi 已提交
532
    root_block = program.block(0)
F
fengjiayi 已提交
533

F
fengjiayi 已提交
534 535
    fwd_op_num = root_block.desc.op_size()
    current_block_idx = program.current_block_idx
F
fengjiayi 已提交
536 537
    grad_to_var = dict()

Y
yuyang18 已提交
538
    op_desc = _create_op_desc_(
X
Xin Pan 已提交
539 540 541 542 543
        "fill_constant",
        {},
        {"Out": [_append_grad_suffix_(loss.name)]},
        {
            "shape": [1],  # TODO(panyx0718): This can be loss.shape.
Y
yuyang18 已提交
544 545 546 547 548 549 550
            "value": 1.0,
            "dtype": loss.dtype,
            "force_cpu": False,
            core.op_proto_and_checker_maker.kOpRoleAttrName():
            int(core.op_proto_and_checker_maker.OpRole.Backward) |
            int(core.op_proto_and_checker_maker.OpRole.Loss),
        })
551 552 553 554
    root_block.desc.append_op().copy_from(op_desc)

    block_no_grad_set = set(map(_strip_grad_suffix_, no_grad_dict[0]))
    op_path = _find_op_path_(root_block, [loss], [], block_no_grad_set)
555 556 557
    no_grad_vars = _find_no_grad_vars(root_block, op_path, [loss],
                                      block_no_grad_set)
    block_no_grad_set.update(no_grad_vars)
558
    no_grad_dict[0].update(list(map(_append_grad_suffix_, block_no_grad_set)))
559

560 561 562 563 564 565 566 567 568 569 570 571 572 573
    input_grad_names_set = None
    # For double backward, input_grad_names is used for filter
    # some non-used gradients op.
    if program._appending_grad_times > 1:
        input_grad_names_set = set([_append_grad_suffix_(loss.name)])

    _append_backward_ops_(
        root_block,
        op_path,
        root_block,
        no_grad_dict,
        grad_to_var,
        callbacks,
        input_grad_names_set=input_grad_names_set)
574 575 576 577 578 579

    # Because calc_gradient may be called multiple times,
    # we need rename the internal gradient variables so that they have
    # different names.
    _rename_grad_(root_block, fwd_op_num, grad_to_var, {})

F
fengjiayi 已提交
580
    _append_backward_vars_(root_block, fwd_op_num, grad_to_var, grad_info_map)
F
fengjiayi 已提交
581

F
fengjiayi 已提交
582
    program.current_block_idx = current_block_idx
W
Wu Yi 已提交
583
    program._sync_with_cpp()
F
fengjiayi 已提交
584

585 586 587
    if parameter_list is not None:
        parameters = parameter_list
    else:
F
fengjiayi 已提交
588
        params = program.global_block().all_parameters()
M
minqiyang 已提交
589
        program.global_block().iter_parameters()
590
        parameters = [param.name for param in params]
591

592 593
    params_and_grads = []
    for param in parameters:
M
minqiyang 已提交
594
        if cpt.to_text(param) not in grad_info_map:
F
fengjiayi 已提交
595
            continue
F
update  
fengjiayi 已提交
596
        grad_info = grad_info_map[param]
F
fengjiayi 已提交
597
        grad_block = grad_info[1]
598 599 600 601
        if not grad_block.has_var(grad_info[0]):
            raise ValueError("grad block[{0}] did not have grad var {1}".format(
                grad_info[1], grad_info[0]))
        # Get the param var from the global block
F
fengjiayi 已提交
602
        param_var = program.global_block().var(param)
603 604 605 606 607
        grad_var = grad_block.var(grad_info[0])
        if loss.block.has_var(grad_info[0]):
            params_and_grads.append((param_var, grad_var))
        else:
            params_and_grads.append((param_var, None))
Y
yuyang18 已提交
608 609 610 611 612 613 614 615 616 617 618 619 620

    op_role_var_attr_name = core.op_proto_and_checker_maker.kOpRoleVarAttrName()
    for p, g in params_and_grads:
        if g is None:
            continue
        for op in reversed(program.global_block().ops):
            assert isinstance(op, framework.Operator)
            if g.name in op.output_arg_names:
                g.op = op
                break

        if g.op is None:
            raise ValueError("Unexpected branch")
Y
yuyang18 已提交
621
        attr_val = [p.name, g.name]
Y
yuyang18 已提交
622 623
        if g.op.has_attr(op_role_var_attr_name):
            attr_val.extend(g.op.attr(op_role_var_attr_name))
W
Wu Yi 已提交
624
        g.op._set_attr(op_role_var_attr_name, attr_val)
Y
yuyang18 已提交
625

626
    return params_and_grads
627 628 629 630 631 632 633 634


def _as_list(x):
    if x is None:
        return []
    return list(x) if isinstance(x, collections.Sequence) else [x]


635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
def _find_no_grad_vars(block, op_path, targets, no_grad_set):
    """
    Find the vars which is not used in the program, and
    those var belong to no_grad_var.
    """
    output_names = set([out.name for out in targets])
    no_grad_var = []
    for i, op in reversed(list(enumerate(op_path))):
        # If the op has sub_block, it is too complicated to find the correct no_grad_var.
        if not op.has_attr("sub_block"):
            for out_var in op.desc.output_arg_names():
                if out_var not in output_names and out_var not in op.desc.input_arg_names(
                ) and not block.vars[out_var].stop_gradient:
                    no_grad_var.append(out_var)
        for name in op.desc.input_arg_names():
            if name not in no_grad_set:
                output_names.add(name)
    return set(no_grad_var)


655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
def _find_op_path_(block, outputs, inputs, no_grad_set):
    """
    no_grad_set will also be changed
    """
    input_names = set([inp.name for inp in inputs])
    output_names = set([out.name for out in outputs])

    relevant_op_flags = [True] * len(block.ops)

    # All the inputs of the block are used if inputs is empty,
    if inputs:
        for i, op in enumerate(block.ops):
            if _some_in_set_(op.desc.input_arg_names(), input_names):
                for name in op.desc.output_arg_names():
                    if name not in no_grad_set:
                        input_names.add(name)
            else:
                relevant_op_flags[i] = False

    for i, op in reversed(list(enumerate(block.ops))):
        if _some_in_set_(op.desc.output_arg_names(), output_names):
            for name in op.desc.input_arg_names():
                if name not in no_grad_set:
                    output_names.add(name)
        else:
            relevant_op_flags[i] = False

    op_path = [
        block.ops[i] for i in range(len(block.ops)) if relevant_op_flags[i]
    ]

    if inputs:
        for op in op_path:
            for name in op.desc.input_arg_names():
689
                if name not in input_names and block.vars[name].stop_gradient:
690 691 692 693 694 695 696
                    no_grad_set.add(name)

    return op_path


def calc_gradient(targets, inputs, target_gradients=None, no_grad_set=None):
    """
697
    Backpropagate the gradients of targets to inputs.
698 699 700 701

    Args:
        targets(Variable|list[Variable]): The target variables
        inputs(Variable|list[Variable]): The input variables
702 703 704
        target_gradients (Variable|list[Variable]|None): The gradient variables
            of targets which has the same shape with targets, If None, ones will
            be created for them.
705 706 707 708 709
        no_grad_set(set[string]): The names of variables that have no gradients
            in Block 0. All variables with `stop_gradient=True` from all blocks
            will be automatically added.

    Return:
710
        (list[Variable]): A list of gradients for inputs
711 712 713 714 715 716 717 718 719
        If an input does not affect targets, the corresponding gradient variable
        will be None
    """
    targets = _as_list(targets)
    inputs = _as_list(inputs)
    target_gradients = _as_list(target_gradients)

    block = targets[0].block
    prog = block.program
720 721
    # increase appending gradients times
    prog._appending_grad_times += 1
722 723 724 725 726 727 728 729 730 731 732 733 734
    block_idx = block.idx

    if not target_gradients:
        target_gradients = [None] * len(targets)

    if len(targets) != len(target_gradients):
        raise ValueError(
            "Should have the same number of target_gradients as targets")

    if no_grad_set is None:
        no_grad_set = set()
    no_grad_set = copy.copy(no_grad_set)
    no_grad_dict = _get_stop_gradients_(prog)
735
    no_grad_dict[0].update(list(map(_append_grad_suffix_, no_grad_set)))
736 737 738

    fwd_op_num = block.desc.op_size()

739 740
    input_grad_names_set = set()

741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
    target_grad_map = {}
    for i, grad in enumerate(target_gradients):
        target = targets[i]
        if grad is None:
            grad_name = _append_grad_suffix_(target.name)
            op_desc = _create_op_desc_("fill_constant_batch_size_like",
                                       {"Input": [target.name]},
                                       {"Out": [grad_name]}, {
                                           "shape": target.shape,
                                           "value": 1.0,
                                           "dtype": target.dtype,
                                           'input_dim_idx': 0,
                                           'output_dim_idx': 0
                                       })
            block.desc.append_op().copy_from(op_desc)
756
            input_grad_names_set.add(grad_name)
757 758 759 760 761 762 763 764
        else:
            if target.block.idx != block_idx or target.block.program != prog:
                raise ValueError("all targets must be in the same block")
            if target.shape != grad.shape:
                raise ValueError(
                    "The shapes of target and grad are different: %s %s" % (
                        target.name, grad.name))
            target_grad_map[_append_grad_suffix_(target.name)] = grad.name
765 766 767 768 769 770
            input_grad_names_set.add(grad.name)

    # For double backward, input_grad_names is used for filter
    # some non-used gradients op.
    if prog._appending_grad_times == 1:
        input_grad_names_set = None
771 772 773 774 775 776 777

    for input in inputs:
        if input.block.program != prog:
            raise "input must be in the same program as targets"

    block_no_grad_set = set(map(_strip_grad_suffix_, no_grad_dict[0]))
    op_path = _find_op_path_(block, targets, inputs, block_no_grad_set)
778
    no_grad_dict[0].update(list(map(_append_grad_suffix_, block_no_grad_set)))
779 780
    grad_to_var = dict()
    grad_info_map = dict()
781 782 783 784 785 786 787
    _append_backward_ops_(
        block,
        op_path,
        block,
        no_grad_dict,
        grad_to_var,
        input_grad_names_set=input_grad_names_set)
788 789 790 791 792 793 794

    # Because calc_gradient may be called multiple times,
    # we need rename the internal gradient variables so that they have
    # different names.
    _rename_grad_(block, fwd_op_num, grad_to_var, target_grad_map)

    _append_backward_vars_(block, fwd_op_num, grad_to_var, grad_info_map)
W
Wu Yi 已提交
795
    prog._sync_with_cpp()
796 797 798 799 800 801 802 803 804 805 806 807 808 809 810

    grad_vars = []
    for input_var in inputs:
        if input_var.name not in grad_info_map:
            grad_vars.append(None)
        else:
            grad_info = grad_info_map[input_var.name]
            grad_block = grad_info[1]
            grad_var = grad_block.var(grad_info[0])
            grad_vars.append(grad_var)

    if len(grad_vars) == 1:
        return grad_vars[0]
    else:
        return grad_vars
811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847


def gradients(targets, inputs, target_gradients=None, no_grad_set=None):
    """
    Backpropagate the gradients of targets to inputs.

    Args:
        targets (Variable|list[Variable]): The target variables.
        inputs (Variable|list[Variable]): The input variables.
        target_gradients (Variable|list[Variable]|None): The gradient variables
            of targets which has the same shape with targets, If None, ones will
            be created for them.
        no_grad_set (set[string]): The names of variables that have no gradients
            in Block 0. All variables with `stop_gradient=True` from all blocks
            will be automatically added.

    Return:
        (list[Variable]): A list of gradients for inputs
        If an input does not affect targets, the corresponding gradient variable
        will be None.

    Examples:
        .. code-block:: python

            import paddle.fluid as fluid

            x = fluid.layers.data(name='x', shape=[2,8,8], dtype='float32')
            x.stop_gradient=False
            y = fluid.layers.conv2d(x, 4, 1, bias_attr=False)
            y = fluid.layers.relu(y)
            y = fluid.layers.conv2d(y, 4, 1, bias_attr=False)
            y = fluid.layers.relu(y)
            z = fluid.gradients([y], x)
            print(z)
    """
    outs = calc_gradient(targets, inputs, target_gradients, no_grad_set)
    return _as_list(outs)