distribute_transpiler.py 54.3 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
# 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.
14 15 16 17 18 19 20 21 22 23 24 25 26
"""
Transpile the program to distributed data-parallelism programs.
The main_program will be transformed to use a remote parameter server
to do parameter optimization. And the optimization graph will be put
into a parameter server program.

Use different methods to split trainable variables to different
parameter servers.

Steps to transpile trainer:
1. split variable to multiple blocks, aligned by product(dim[1:]) (width).
2. rename splited grad variables to add trainer_id suffix ".trainer_%d".
3. modify trainer program add split_op to each grad variable.
27 28 29
4. append send_op to send splited variables to server and 
5. add recv_op to fetch params(splited blocks or origin param) from server.
6. append concat_op to merge splited blocks to update local weights.
30 31 32 33 34 35 36 37

Steps to transpile pserver:
1. create new program for parameter server.
2. create params and grad variables that assigned to current server instance.
3. create a sub-block in the server side program
4. append ops that should run on current server instance.
5. add listen_and_serv op
"""
D
dzhwinter 已提交
38

T
typhoonzero 已提交
39
from __future__ import print_function
40

T
typhoonzero 已提交
41
import math
42
import numpy as np
43

Y
Yancey1989 已提交
44
from ps_dispatcher import RoundRobin, HashName, PSDispatcher
Y
Yancey 已提交
45
from .. import core, framework
T
typhoonzero 已提交
46 47 48
from ..framework import Program, default_main_program, \
                        default_startup_program, \
                        Variable, Parameter, grad_var_name
49
from details import *
50 51 52

LOOKUP_TABLE_TYPE = "lookup_table"
LOOKUP_TABLE_GRAD_TYPE = "lookup_table_grad"
53
OP_ROLE_VAR_ATTR_NAME = core.op_proto_and_checker_maker.kOpRoleVarAttrName()
Y
Yancey1989 已提交
54 55 56
RPC_OP_ROLE_ATTR_NAME = op_role_attr_name = core.op_proto_and_checker_maker.kOpRoleAttrName(
)
RPC_OP_ROLE_ATTR_VALUE = core.op_proto_and_checker_maker.OpRole.RPC
T
done  
typhoonzero 已提交
57 58


T
typhoonzero 已提交
59 60 61 62 63 64
class VarBlock:
    def __init__(self, varname, offset, size):
        self.varname = varname
        # NOTE: real offset is offset * size
        self.offset = offset
        self.size = size
T
done  
typhoonzero 已提交
65

T
typhoonzero 已提交
66 67
    def __str__(self):
        return "%s:%d:%d" % (self.varname, self.offset, self.size)
T
done  
typhoonzero 已提交
68 69


70 71 72 73
def same_or_split_var(p_name, var_name):
    return p_name == var_name or p_name.startswith(var_name + ".block")


74
def slice_variable(var_list, slice_count, min_block_size=8192):
T
typhoonzero 已提交
75
    """
76 77 78 79 80 81
    We may need to split dense tensor to one or more blocks and put
    them equally onto parameter server. One block is a sub-tensor
    aligned by dim[0] of the tensor.

    We need to have a minimal block size so that the calculations in
    the parameter server side can gain better performance. By default
82
    minimum block size 8K elements (maybe 16bit or 32bit or 64bit).
83 84 85

    Args:
        var_list (list): List of variables.
86 87
        slice_count (int): Numel of count that variables will be sliced, which
            could be the pserver services' count.
88 89
        min_block_size (int): Minimum splitted block size.
    Returns:
90
        blocks (list[(varname, block_id, current_block_size)]): A list
91
            of VarBlocks. Each VarBlock specifies a shard of the var.
T
typhoonzero 已提交
92 93 94
    """
    blocks = []
    for var in var_list:
95
        split_count = slice_count
T
typhoonzero 已提交
96 97 98 99
        var_numel = reduce(lambda x, y: x * y, var.shape)
        max_pserver_count = int(math.floor(var_numel / float(min_block_size)))
        if max_pserver_count == 0:
            max_pserver_count = 1
100
        if max_pserver_count < slice_count:
T
typhoonzero 已提交
101 102 103 104 105 106 107 108 109
            split_count = max_pserver_count
        block_size = int(math.ceil(var_numel / float(split_count)))

        if len(var.shape) >= 2:
            # align by dim1(width)
            dim1 = reduce(lambda x, y: x * y, var.shape[1:])
            remains = block_size % dim1
            if remains != 0:
                block_size += dim1 - remains
110
        # update split_count after aligning
T
typhoonzero 已提交
111 112 113 114 115 116 117 118 119
        split_count = int(math.ceil(var_numel / float(block_size)))
        for block_id in xrange(split_count):
            curr_block_size = min(block_size, var_numel - (
                (block_id) * block_size))
            block = VarBlock(var.name, block_id, curr_block_size)
            blocks.append(str(block))
    return blocks


T
done  
typhoonzero 已提交
120
class DistributeTranspiler:
121
    def _has_distributed_lookup_table(self):
122 123 124 125 126 127
        # process lookup_table_op
        # 1. check all lookup_table_op is distributed
        # 2. check all lookup_table_op share the same table.
        distributed_lookup_table_ops = []
        # support only one distributed_lookup_table now
        self.table_name = None
128
        for op in self.origin_program.global_block().ops:
129 130 131 132 133 134 135 136 137 138 139 140
            if op.type == LOOKUP_TABLE_TYPE:
                if op.attrs['is_distributed'] is True:
                    if self.table_name is None:
                        self.table_name = op.input("W")[0]
                    if self.table_name != op.input("W")[0]:
                        raise RuntimeError("all distributed lookup_table_ops"
                                           " should have only one table")
                    distributed_lookup_table_ops.append(op)
                else:
                    if self.table_name is not None:
                        assert op.input("W")[0] != self.table_name

141
        return len(distributed_lookup_table_ops) > 0
142

143 144 145 146 147
    def _update_dist_lookup_table_vars(self, param_list, grad_list,
                                       params_grads):
        # TODO(wuyi): put find a way to put dist lookup table stuff all together.
        # update self.table_param_grad and self.trainer_side_table_grad_list
        program = self.origin_program
148 149 150 151 152 153
        if self.has_distributed_lookup_table:
            param_list = [
                param for param in param_list if param.name != self.table_name
            ]
            grad_list = [
                grad for grad in grad_list
T
typhoonzero 已提交
154
                if grad.name != grad_var_name(self.table_name)
155 156 157 158 159 160
            ]
            self.table_param_grad = [
                param_grad for param_grad in params_grads
                if param_grad[0].name == self.table_name
            ][0]
            table_grad_var = self.table_param_grad[1]
161
            if self.sync_mode:
162
                self.trainer_side_table_grad_list = [
163 164
                    program.global_block().create_var(
                        name="%s.trainer_%d.pserver_%d" %
165
                        (table_grad_var.name, self.trainer_id, index),
166 167 168 169 170 171
                        type=table_grad_var.type,
                        shape=table_grad_var.shape,
                        dtype=table_grad_var.dtype)
                    for index in range(len(self.pserver_endpoints))
                ]
            else:
172
                self.trainer_side_table_grad_list = [
173 174 175 176 177 178 179
                    program.global_block().create_var(
                        name="%s.pserver_%d" % (table_grad_var.name, index),
                        type=table_grad_var.type,
                        shape=table_grad_var.shape,
                        dtype=table_grad_var.dtype)
                    for index in range(len(self.pserver_endpoints))
                ]
Q
qiaolongfei 已提交
180
        return param_list, grad_list
181

182
    def _init_splited_vars(self, slice_var_up):
183 184 185 186 187 188 189 190
        # update these mappings for further transpile:
        # 1. param_var_mapping: param var name -> [splited params vars]
        # 2. grad_var_mapping: grad var name -> [splited grads vars]
        # 3. grad_param_mapping: grad.blockx -> param.blockx
        # 4. param_grad_ep_mapping: ep -> {"params": [], "grads": []}

        param_list = []
        grad_list = []
Y
yi.wu 已提交
191
        param_grad_set = set()
192 193 194 195
        for p, g in self.params_grads:
            # skip parameter marked not trainable
            if type(p) == Parameter and p.trainable == False:
                continue
Y
yi.wu 已提交
196 197 198 199 200 201
            if p.name not in param_grad_set:
                param_list.append(p)
                param_grad_set.add(p.name)
            if g.name not in param_grad_set:
                grad_list.append(g)
                param_grad_set.add(g.name)
202

Q
qiaolongfei 已提交
203 204
        param_list, grad_list = self._update_dist_lookup_table_vars(
            param_list, grad_list, self.params_grads)
205

206 207 208 209 210
        if slice_var_up:
            # when we slice var up into blocks, we will slice the var according to
            # pserver services' count. A pserver may have two or more listening ports.
            grad_blocks = slice_variable(grad_list, len(self.pserver_endpoints))
            param_blocks = slice_variable(param_list,
211
                                          len(self.pserver_endpoints))
212
        else:
213
            # when we do NOT slice var up into blocks, we will always slice params
214
            # grads into one block.
215 216
            grad_blocks = slice_variable(grad_list, 1)
            param_blocks = slice_variable(param_list, 1)
Y
update  
Yancey1989 已提交
217
        assert (len(grad_blocks) == len(param_blocks))
218

219 220 221 222 223 224 225 226
        # origin_varname -> [splited_var]
        self.param_var_mapping = self._create_vars_from_blocklist(
            self.origin_program, param_blocks)
        self.grad_var_mapping = self._create_vars_from_blocklist(
            self.origin_program,
            grad_blocks,
            add_trainer_suffix=self.trainer_num > 1)
        self.grad_param_mapping = dict()
Y
update  
Yancey1989 已提交
227 228 229
        for g, p in zip(grad_blocks, param_blocks):
            g_name, g_bid, _ = g.split(":")
            p_name, p_bid, _ = p.split(":")
230 231
            self.grad_param_mapping[self.grad_var_mapping[g_name][int(g_bid)]] =  \
                    self.param_var_mapping[p_name][int(p_bid)]
232

233
        # create mapping of endpoint -> split var to create pserver side program
T
typhoonzero 已提交
234
        self.param_grad_ep_mapping = dict()
Y
Yancey1989 已提交
235 236 237 238 239 240 241 242 243
        [
            self.param_grad_ep_mapping.update({
                ep: {
                    "params": [],
                    "grads": []
                }
            }) for ep in self.pserver_endpoints
        ]

244 245 246 247 248
    def transpile(self,
                  trainer_id,
                  program=None,
                  pservers="127.0.0.1:6174",
                  trainers=1,
249
                  slice_var_up=True,
250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
                  split_method=RoundRobin,
                  sync_mode=True):
        """
        :param trainer_id: one unique id for each trainer in a job.
        :type trainer_id: int
        :param program: program to transpile, default is default_main_program
        :type program: Program
        :param pservers: parameter server endpoints like "m1:6174,m2:6174"
        :type pservers: string
        :param trainers: total number of workers/trainers in the job
        :type trainers: int
        :param split_method: A function to determin how to split variables
            to different servers equally.
        :type split_method: function
        :type sync_mode: boolean default True
        :param sync_mode: if sync_mode is set True, it means that dist transpiler
        will transpile the program into sync_mode pserver and trainer program.
        """
        assert (split_method.__bases__[0] == PSDispatcher)
        if program is None:
            program = default_main_program()
        self.origin_program = program
        self.trainer_num = trainers
        self.sync_mode = sync_mode
        self.trainer_id = trainer_id
        pserver_endpoints = pservers.split(",")
        self.pserver_endpoints = pserver_endpoints
        self.optimize_ops, self.params_grads = self._get_optimize_pass()

        ps_dispatcher = split_method(self.pserver_endpoints)
        self.has_distributed_lookup_table = self._has_distributed_lookup_table()

        # split and create vars, then put splited vars in dicts for later use.
283
        self._init_splited_vars(slice_var_up)
284

Y
Yancey1989 已提交
285 286
        # step 3.1: insert send op to send gradient vars to parameter servers
        ps_dispatcher.reset()
Y
update  
Yancey1989 已提交
287
        send_vars = []
288 289 290 291 292 293

        # in general cases, the number of pservers is times of 2, and this
        # will lead to uneven distribution among weights and bias:
        #       fc_w@GRAD_trainer_0, fc_w@GRAD_trainer_1 --> pserver1
        #       fc_b@GRAD_trainer_0, fc_b@GRAD_trainer_1 --> pserver2
        # shuffle the map will avoid the uneven distribution above
294
        grad_var_mapping_items = self.grad_var_mapping.items()
295
        if not slice_var_up:
296 297 298
            np.random.shuffle(grad_var_mapping_items)

        for orig_varname, splited_vars in grad_var_mapping_items:
Y
update  
Yancey1989 已提交
299
            eplist = ps_dispatcher.dispatch(splited_vars)
300

301
            if not slice_var_up:
302 303
                assert (len(splited_vars) == 1)

Y
Yancey1989 已提交
304 305 306 307 308 309 310 311 312
            if len(splited_vars) == 1:
                orig_varname = splited_vars[0].name
                index = find_op_by_output_arg(program.global_block(),
                                              orig_varname)
            elif len(splited_vars) > 1:
                orig_var = program.global_block().vars[orig_varname]
                index = find_op_by_output_arg(program.global_block(),
                                              orig_varname)
                self._insert_split_op(program, orig_var, index, splited_vars)
Y
update  
Yancey1989 已提交
313
                index += 1
Y
Yancey1989 已提交
314 315 316 317
            else:
                AssertionError("Can not insert the send op by original "
                               "variable name :", orig_varname)

Y
Yancey1989 已提交
318
            program.global_block().insert_op(
Y
update  
Yancey1989 已提交
319
                index=index + 1,
320
                type="send",
Y
update  
Yancey1989 已提交
321
                inputs={"X": splited_vars},
Y
Yancey1989 已提交
322 323 324 325 326
                outputs={},
                attrs={
                    "epmap": eplist,
                    RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
                })
Y
update  
Yancey1989 已提交
327 328
            for _, var in enumerate(splited_vars):
                send_vars.append(var)
Y
Yancey1989 已提交
329 330 331 332 333

        if self.sync_mode:
            program.global_block().append_op(
                type="send_barrier",
                inputs={},
Y
Yancey1989 已提交
334
                outputs={},
Y
Yancey1989 已提交
335 336
                attrs={
                    "endpoints": pserver_endpoints,
Y
Yancey1989 已提交
337 338
                    "sync_mode": self.sync_mode,
                    RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
Y
Yancey1989 已提交
339
                })
Y
Yancey1989 已提交
340 341 342

        # step 3.2: insert recv op to receive parameters from parameter server
        recv_vars = []
Y
update  
Yancey1989 已提交
343
        for _, var in enumerate(send_vars):
344
            recv_vars.append(self.grad_param_mapping[var])
Y
update  
Yancey1989 已提交
345
        ps_dispatcher.reset()
Y
Yancey1989 已提交
346 347
        eplist = ps_dispatcher.dispatch(recv_vars)

T
typhoonzero 已提交
348
        for i, ep in enumerate(eplist):
Y
Yancey1989 已提交
349 350
            self.param_grad_ep_mapping[ep]["params"].append(recv_vars[i])
            self.param_grad_ep_mapping[ep]["grads"].append(send_vars[i])
351

Y
Yancey1989 已提交
352
        # step4: Concat the parameters splits together after recv.
353
        for varname, splited_var in self.param_var_mapping.iteritems():
Y
Yancey1989 已提交
354 355 356 357 358 359 360 361
            eps = []
            for var in splited_var:
                index = [v.name for v in recv_vars].index(var.name)
                eps.append(eplist[index])

            program.global_block().append_op(
                type="recv",
                inputs={},
Y
Yancey1989 已提交
362 363 364 365 366
                outputs={"Out": splited_var},
                attrs={
                    "epmap": eps,
                    RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
                })
T
typhoonzero 已提交
367

T
typhoonzero 已提交
368
        program.global_block().append_op(
Y
Yancey1989 已提交
369 370
            type="fetch_barrier",
            inputs={},
Y
Yancey1989 已提交
371
            outputs={},
Q
qiaolongfei 已提交
372 373
            attrs={
                "endpoints": pserver_endpoints,
Y
Yancey1989 已提交
374
                RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
Q
qiaolongfei 已提交
375
            })
Y
Yancey1989 已提交
376

377
        for varname, splited_var in self.param_var_mapping.iteritems():
T
typhoonzero 已提交
378 379
            if len(splited_var) <= 1:
                continue
T
typhoonzero 已提交
380
            orig_param = program.global_block().vars[varname]
T
typhoonzero 已提交
381
            program.global_block().append_op(
T
typhoonzero 已提交
382
                type="concat",
T
typhoonzero 已提交
383
                inputs={"X": splited_var},
T
typhoonzero 已提交
384
                outputs={"Out": [orig_param]},
T
typhoonzero 已提交
385
                attrs={"axis": 0})
T
typhoonzero 已提交
386

387
        if self.has_distributed_lookup_table:
Q
update  
qiaolongfei 已提交
388 389
            self._replace_lookup_table_op_with_prefetch(program,
                                                        pserver_endpoints)
Y
Yancey1989 已提交
390
            self._split_table_grad_and_add_send_vars(program, pserver_endpoints)
391

T
typhoonzero 已提交
392 393
    def get_trainer_program(self):
        # remove optimize ops and add a send op to main_program
394
        delete_ops(self.origin_program.global_block(), self.optimize_ops)
395
        # FIXME(typhoonzero): serialize once will fix error occurs when clone.
396 397
        self.origin_program.__str__()
        return self.origin_program
T
typhoonzero 已提交
398 399 400 401

    def get_pserver_program(self, endpoint):
        """
        Get pserver side program using the endpoint.
402
        TODO(panyx0718): Revisit this assumption. what if #blocks > #pservers.
T
typhoonzero 已提交
403 404 405 406 407 408
        NOTE: assume blocks of the same variable is not distributed
        on the same pserver, only change param/grad varnames for
        trainers to fetch.
        """
        # step1
        pserver_program = Program()
409
        # step2: Create vars to receive vars at parameter servers.
T
typhoonzero 已提交
410 411 412 413 414 415 416 417
        recv_inputs = []
        for v in self.param_grad_ep_mapping[endpoint]["params"]:
            self._clone_var(pserver_program.global_block(), v)
        for v in self.param_grad_ep_mapping[endpoint]["grads"]:
            # create vars for each trainer in global scope, so
            # we don't need to create them when grad arrives.
            # change client side var name to origin name by
            # removing ".trainer_%d" suffix
T
update  
typhoonzero 已提交
418 419 420 421 422
            suff_idx = v.name.find(".trainer_")
            if suff_idx >= 0:
                orig_var_name = v.name[:suff_idx]
            else:
                orig_var_name = v.name
T
typhoonzero 已提交
423 424 425 426 427 428 429 430 431
            # NOTE: single_trainer_var must be created for multi-trainer
            # case to merge grads from multiple trainers
            single_trainer_var = \
                pserver_program.global_block().create_var(
                    name=orig_var_name,
                    persistable=True,
                    type=v.type,
                    dtype=v.dtype,
                    shape=v.shape)
432
            if self.sync_mode and self.trainer_num > 1:
433
                for trainer_id in xrange(self.trainer_num):
T
typhoonzero 已提交
434 435 436 437 438 439 440 441 442
                    var = pserver_program.global_block().create_var(
                        name="%s.trainer_%d" % (orig_var_name, trainer_id),
                        persistable=False,
                        type=v.type,
                        dtype=v.dtype,
                        shape=v.shape)
                    recv_inputs.append(var)
            else:
                recv_inputs.append(single_trainer_var)
443

Q
qiaolongfei 已提交
444
        # step 3
445
        # Create a union-find data structure from optimize ops,
T
typhoonzero 已提交
446 447 448
        # If two ops are connected, we could add these two ops
        # into one set.
        ufind = self._create_ufind(self.optimize_ops)
Q
qiaolongfei 已提交
449
        # step 3.2
T
typhoonzero 已提交
450 451 452 453
        # Iterate through the ops and append optimize op which
        # located on current pserver
        opt_op_on_pserver = []
        for _, op in enumerate(self.optimize_ops):
454 455
            if self._is_optimizer_op(op) and self._is_opt_op_on_pserver(
                    endpoint, op):
T
typhoonzero 已提交
456
                opt_op_on_pserver.append(op)
Q
qiaolongfei 已提交
457
        # step 3.3
T
typhoonzero 已提交
458
        # Iterate through the ops, and if an op and the optimize ops
459
        # which located on current pserver are in one set, then
T
typhoonzero 已提交
460
        # append it into the sub program.
T
typhoonzero 已提交
461 462 463 464 465

        global_ops = []
        # HACK: optimization global ops only used to scale beta1 and beta2
        # replace it with dependency engine.
        for op in self.optimize_ops:
466 467
            if self._is_adam_connected_op(op):
                global_ops.append(op)
T
typhoonzero 已提交
468

469 470
        def __append_optimize_op__(op, block, grad_to_block_id, merged_var):
            if self._is_optimizer_op(op):
Q
qiaolongfei 已提交
471
                self._append_pserver_ops(block, op, endpoint, grad_to_block_id,
472
                                         self.origin_program, merged_var)
T
typhoonzero 已提交
473
            else:
474 475 476 477 478 479 480
                self._append_pserver_non_opt_ops(block, op, endpoint)

        def __op_have_grad_input__(op):
            for varname in op.input_arg_names:
                if varname.find("@GRAD") >= 0:
                    return varname
            return ""
T
typhoonzero 已提交
481

482
        # append lr decay ops to the child block if exists
483 484
        lr_ops = self._get_lr_ops()
        if len(lr_ops) > 0:
Q
qiaolongfei 已提交
485 486
            lr_decay_block = pserver_program.create_block(
                pserver_program.num_blocks - 1)
487
            for _, op in enumerate(lr_ops):
488
                self._append_pserver_non_opt_ops(lr_decay_block, op, endpoint)
489

T
typhoonzero 已提交
490
        # append op to the current block
Q
qiaolongfei 已提交
491
        grad_to_block_id = []
Q
qiaolongfei 已提交
492
        pre_block_idx = pserver_program.num_blocks - 1
T
typhoonzero 已提交
493
        for idx, opt_op in enumerate(opt_op_on_pserver):
494
            per_opt_block = pserver_program.create_block(pre_block_idx)
495 496 497 498 499 500 501 502
            # append grad merging ops before clip and weight decay
            for _, op in enumerate(self.optimize_ops):
                # find the origin @GRAD var before clipping
                grad_varname_for_block = __op_have_grad_input__(op)
                if ufind.is_connected(op, opt_op) and grad_varname_for_block:
                    merged_var = self._append_pserver_grad_merge_ops(
                        per_opt_block, grad_varname_for_block, endpoint,
                        grad_to_block_id, self.origin_program)
T
typhoonzero 已提交
503 504
            for _, op in enumerate(self.optimize_ops):
                # optimizer is connected to itself
505
                if ufind.is_connected(op, opt_op) and op not in global_ops:
506 507
                    __append_optimize_op__(op, per_opt_block, grad_to_block_id,
                                           merged_var)
T
typhoonzero 已提交
508 509

        # append global ops
510
        if global_ops:
Q
qiaolongfei 已提交
511 512 513
            opt_state_block = pserver_program.create_block(
                pserver_program.num_blocks - 1)
            for glb_op in global_ops:
X
Xi Chen 已提交
514
                __append_optimize_op__(glb_op, opt_state_block,
515
                                       grad_to_block_id, None)
T
typhoonzero 已提交
516

517 518 519 520
        # process distributed lookup_table
        prefetch_block = None
        if self.has_distributed_lookup_table:
            pserver_index = self.pserver_endpoints.index(endpoint)
521
            table_opt_block = self._create_table_optimize_block(
522
                pserver_index, pserver_program, pre_block_idx, grad_to_block_id)
523
            prefetch_block = self._create_prefetch_block(
524
                pserver_index, pserver_program, table_opt_block)
525 526 527 528 529 530 531 532 533

        # NOTE: if has_distributed_lookup_table is False, then prefetch_block will
        # not be executed, so it's safe to use optimize_block to hold the place
        if self.has_distributed_lookup_table:
            assert prefetch_block is not None
        else:
            assert prefetch_block is None
            prefetch_block = pserver_program.global_block()

T
typhoonzero 已提交
534 535 536 537 538 539
        # step5 append the listen_and_serv op
        pserver_program.global_block().append_op(
            type="listen_and_serv",
            inputs={'X': recv_inputs},
            outputs={},
            attrs={
Q
qiaolongfei 已提交
540
                "OptimizeBlock": pserver_program.block(1),
T
typhoonzero 已提交
541
                "endpoint": endpoint,
542
                "Fanin": self.trainer_num,
Q
tmp  
qiaolongfei 已提交
543 544
                "PrefetchBlock": prefetch_block,
                "sync_mode": self.sync_mode,
Q
qiaolongfei 已提交
545
                "grad_to_block_id": grad_to_block_id
T
typhoonzero 已提交
546
            })
547

T
typhoonzero 已提交
548 549 550 551 552 553 554 555 556 557
        pserver_program.sync_with_cpp()
        return pserver_program

    def get_startup_program(self, endpoint, pserver_program):
        """
        Get startup program for current parameter server.
        Modify operator input variables if there are variables that
        were split to several blocks.
        """
        s_prog = Program()
T
typhoonzero 已提交
558
        orig_s_prog = default_startup_program()
T
typhoonzero 已提交
559 560 561 562 563 564 565 566 567 568 569 570 571
        params = self.param_grad_ep_mapping[endpoint]["params"]

        def _get_splited_name_and_shape(varname):
            for idx, splited_param in enumerate(params):
                pname = splited_param.name
                if same_or_split_var(pname, varname) and varname != pname:
                    return pname, splited_param.shape
            return "", []

        # 1. create vars in pserver program to startup program
        pserver_vars = pserver_program.global_block().vars
        created_var_map = dict()
        for _, var in pserver_vars.iteritems():
T
update  
typhoonzero 已提交
572
            tmpvar = s_prog.global_block().clone_variable(var)
T
typhoonzero 已提交
573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604
            created_var_map[var.name] = tmpvar

        # 2. rename op outputs
        for op in orig_s_prog.global_block().ops:
            new_inputs = dict()
            new_outputs = dict()
            # do not append startup op if var is not on this pserver
            op_on_pserver = False
            for key in op.output_names:
                newname, _ = _get_splited_name_and_shape(op.output(key)[0])
                if newname:
                    op_on_pserver = True
                    new_outputs[key] = created_var_map[newname]
                elif op.output(key)[0] in pserver_vars:
                    op_on_pserver = True
                    new_outputs[key] = pserver_vars[op.output(key)[0]]

            # most startup program ops have no inputs
            new_inputs = self._get_input_map_from_op(pserver_vars, op)

            if op_on_pserver:
                if op.type in [
                        "gaussian_random", "fill_constant", "uniform_random"
                ]:
                    op.attrs["shape"] = new_outputs["Out"].shape
                s_prog.global_block().append_op(
                    type=op.type,
                    inputs=new_inputs,
                    outputs=new_outputs,
                    attrs=op.attrs)
        return s_prog

605 606
    # ====================== private transpiler functions =====================

607
    # transpiler function for dis lookup_table
Q
update  
qiaolongfei 已提交
608 609
    def _replace_lookup_table_op_with_prefetch(self, program,
                                               pserver_endpoints):
610 611 612 613 614 615 616 617 618 619 620 621
        # 1. replace lookup_table_op with split_ids_op -> prefetch_op -> sum_op
        self.prefetch_input_vars = None
        self.prefetch_output_vars = None

        continue_search_lookup_table_op = True
        while continue_search_lookup_table_op:
            continue_search_lookup_table_op = False
            all_ops = program.global_block().ops
            for op in all_ops:
                if op.type == LOOKUP_TABLE_TYPE:
                    continue_search_lookup_table_op = True

622
                    lookup_table_op_index = list(all_ops).index(op)
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
                    ids_name = op.input("Ids")
                    out_name = op.output("Out")

                    if self.prefetch_input_vars is None:
                        ids_var = program.global_block().vars[ids_name[0]]
                        self.prefetch_input_vars = self.create_splited_vars(
                            source_var=ids_var,
                            block=program.global_block(),
                            tag="_prefetch_in_")
                    if self.prefetch_output_vars is None:
                        out_var = program.global_block().vars[out_name[0]]
                        self.prefetch_output_vars = self.create_splited_vars(
                            source_var=out_var,
                            block=program.global_block(),
                            tag="_prefetch_out_")

                    # insert split_ids_op
                    program.global_block().insert_op(
641
                        index=lookup_table_op_index,
642 643 644 645 646 647 648 649 650 651 652
                        type="split_ids",
                        inputs={
                            'Ids': [
                                program.global_block().vars[varname]
                                for varname in ids_name
                            ]
                        },
                        outputs={"Out": self.prefetch_input_vars})

                    # insert prefetch_op
                    program.global_block().insert_op(
653
                        index=lookup_table_op_index + 1,
654 655
                        type="prefetch",
                        inputs={'X': self.prefetch_input_vars},
Y
Yancey1989 已提交
656 657
                        outputs={"Out": self.prefetch_output_vars},
                        attrs={
658
                            "epmap": pserver_endpoints,
Y
Yancey1989 已提交
659 660
                            RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
                        })
661 662 663

                    # insert concat_op
                    program.global_block().insert_op(
664 665 666 667 668 669 670 671 672
                        index=lookup_table_op_index + 2,
                        type="merge_ids",
                        inputs={
                            'Ids': [
                                program.global_block().vars[varname]
                                for varname in ids_name
                            ],
                            'X': self.prefetch_output_vars
                        },
673 674 675 676 677
                        outputs={
                            "Out": [
                                program.global_block().vars[varname]
                                for varname in out_name
                            ]
678
                        })
679 680

                    # delete lookup_table_op
681
                    delete_ops(program.global_block(), [op])
682 683 684
                    # break for loop
                    break

Y
Yancey1989 已提交
685
    def _split_table_grad_and_add_send_vars(self, program, pserver_endpoints):
686
        # 2. add split_ids_op and send_op to send gradient to pservers
687 688
        # there should only be one table_name
        all_ops = program.global_block().ops
T
typhoonzero 已提交
689
        table_grad_name = grad_var_name(self.table_name)
690 691 692 693 694 695 696 697 698 699
        for op in all_ops:
            if table_grad_name in op.output_arg_names:
                op_index = list(all_ops).index(op)
                # insert split_ids_op
                program.global_block().insert_op(
                    index=op_index + 1,
                    type="split_ids",
                    inputs={
                        'Ids': [program.global_block().vars[table_grad_name]]
                    },
700
                    outputs={"Out": self.trainer_side_table_grad_list})
701 702
                program.global_block().insert_op(
                    index=op_index + 2,
703
                    type="send",
704
                    inputs={'X': self.trainer_side_table_grad_list},
Y
Yancey1989 已提交
705 706
                    outputs={},
                    attrs={
707
                        "sync_mode": True,
Y
Yancey1989 已提交
708 709 710
                        "epmap": pserver_endpoints,
                        RPC_OP_ROLE_ATTR_NAME: RPC_OP_ROLE_ATTR_VALUE
                    })
711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730
                break

    def _create_prefetch_block(self, pserver_index, pserver_program,
                               optimize_block):
        # STEP: create prefetch block
        table_var = pserver_program.global_block().vars[self.table_name]
        prefetch_block = pserver_program.create_block(optimize_block.idx)
        trainer_ids = self.prefetch_input_vars[pserver_index]
        pserver_ids = pserver_program.global_block().create_var(
            name=trainer_ids.name,
            type=trainer_ids.type,
            shape=trainer_ids.shape,
            dtype=trainer_ids.dtype)
        trainer_out = self.prefetch_output_vars[pserver_index]
        pserver_out = pserver_program.global_block().create_var(
            name=trainer_out.name,
            type=trainer_out.type,
            shape=trainer_out.shape,
            dtype=trainer_out.dtype)
        prefetch_block.append_op(
Y
Yancey1989 已提交
731
            type="lookup_sparse_table",
732 733 734 735 736 737 738 739 740 741 742
            inputs={'Ids': pserver_ids,
                    "W": table_var},
            outputs={"Out": pserver_out},
            attrs={
                "is_sparse": True,  # has no effect on lookup_table op
                "is_distributed": True,
                "padding_idx": -1
            })
        return prefetch_block

    def _create_table_optimize_block(self, pserver_index, pserver_program,
743
                                     pre_block_idx, grad_to_block_id):
744 745
        # STEP: create table optimize block
        # create table param and grad var in pserver program
Y
Yancey1989 已提交
746 747 748 749 750 751 752 753
        origin_param_var = self.origin_program.global_block().vars[
            self.table_name]
        param_var = pserver_program.global_block().create_var(
            name=origin_param_var.name,
            shape=origin_param_var.shape,
            dtype=origin_param_var.dtype,
            type=core.VarDesc.VarType.SELECTED_ROWS,
            persistable=True)
754 755 756
        # parameter must be selected rows
        param_var.desc.set_type(core.VarDesc.VarType.SELECTED_ROWS)
        grad_var = pserver_program.global_block().clone_variable(
T
typhoonzero 已提交
757
            self.origin_program.global_block().vars[grad_var_name(
758
                self.table_name)])
759 760 761 762 763 764

        # create table optimize block in pserver program
        table_opt_op = [
            op for op in self.optimize_ops
            if op.input("Param")[0] == self.table_name
        ][0]
Q
qiaolongfei 已提交
765
        table_opt_block = pserver_program.create_block(pre_block_idx)
766 767 768
        # only support sgd now
        assert table_opt_op.type == "sgd"

769 770 771
        if self.sync_mode:
            # create grad vars in pserver program
            table_grad_var = self.table_param_grad[1]
772
            pserver_side_table_grad_list = [
773 774 775 776 777 778 779 780 781
                pserver_program.global_block().create_var(
                    name="%s.trainer_%d.pserver_%d" %
                    (table_grad_var.name, index, pserver_index),
                    type=table_grad_var.type,
                    shape=table_grad_var.shape,
                    dtype=table_grad_var.dtype)
                for index in range(self.trainer_num)
            ]

782
            # append sum op for pserver_side_table_grad_list
783 784
            table_opt_block.append_op(
                type="sum",
785
                inputs={"X": pserver_side_table_grad_list},
786
                outputs={"Out": [grad_var]})
787 788
        else:
            # in async_mode, for table gradient, it also need to be splited to each parameter server
789
            origin_grad_name = grad_var.name
790 791
            splited_grad_name = self.trainer_side_table_grad_list[
                pserver_index].name
792 793
            if not splited_grad_name.startswith(origin_grad_name):
                raise ValueError("origin_grad_var: " + splited_grad_name +
794
                                 " grad_var:" + grad_var.name)
795 796
            grad_var = pserver_program.global_block().rename_var(
                origin_grad_name, splited_grad_name)
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811

        lr_var = pserver_program.global_block().vars[table_opt_op.input(
            "LearningRate")[0]]
        inputs = {
            "Param": [param_var],
            "Grad": [grad_var],
            "LearningRate": [lr_var]
        }
        outputs = {"ParamOut": [param_var]}
        table_opt_block.append_op(
            type=table_opt_op.type,
            inputs=inputs,
            outputs=outputs,
            attrs=table_opt_op.attrs)

812 813 814
        # add table parameter gradient and it's block id to grad_to_block_id
        grad_to_block_id.append(grad_var.name + ":" + str(table_opt_block.idx))

815 816
        return table_opt_block

T
typhoonzero 已提交
817 818 819 820 821
    def _create_vars_from_blocklist(self,
                                    program,
                                    block_list,
                                    add_trainer_suffix=False):
        """
822
        Create vars for each split.
T
typhoonzero 已提交
823 824
        NOTE: only grads need to be named for different trainers, use
              add_trainer_suffix to rename the grad vars.
825 826 827 828
        Args:
            program (ProgramDesc): ProgramDesc which gradients blong.
            block_list (list[(varname, block_id, block_size)]): List of gradient blocks.
            add_trainer_suffix (Bool): Add trainer suffix to new variable's name if set True.
829 830
        Returns:
            var_mapping (dict(varname->[new_varname_variable])):A dict mapping
831
                from original var name to each var split.
T
typhoonzero 已提交
832
        """
833 834

        # varname->[(block_id, current_block_size)]
T
typhoonzero 已提交
835
        block_map = dict()
836

T
typhoonzero 已提交
837
        var_mapping = dict()
T
typhoonzero 已提交
838 839 840 841 842
        for block_str in block_list:
            varname, offset, size = block_str.split(":")
            if not block_map.has_key(varname):
                block_map[varname] = []
            block_map[varname].append((long(offset), long(size)))
Y
yi.wu 已提交
843 844 845
        # Do not remove this important debug message:
        print("block map: %s" % block_map)

T
typhoonzero 已提交
846
        for varname, splited in block_map.iteritems():
T
typhoonzero 已提交
847
            orig_var = program.global_block().var(varname)
T
typhoonzero 已提交
848
            if len(splited) == 1:
849
                if self.sync_mode and add_trainer_suffix:
T
typhoonzero 已提交
850 851 852 853 854 855 856 857
                    new_var_name = "%s.trainer_%d" % \
                        (orig_var.name, self.trainer_id)
                    program.global_block().rename_var(varname, new_var_name)
                    var_mapping[varname] = \
                        [program.global_block().var(new_var_name)]
                else:
                    var_mapping[varname] = \
                        [program.global_block().var(orig_var.name)]
T
typhoonzero 已提交
858
                continue
T
typhoonzero 已提交
859 860

            var_mapping[varname] = []
T
typhoonzero 已提交
861 862 863 864
            orig_shape = orig_var.shape
            orig_dim1_flatten = 1
            if len(orig_shape) >= 2:
                orig_dim1_flatten = reduce(lambda x, y: x * y, orig_shape[1:])
T
typhoonzero 已提交
865

T
typhoonzero 已提交
866
            for i, block in enumerate(splited):
T
typhoonzero 已提交
867
                size = block[1]
T
typhoonzero 已提交
868 869 870 871
                rows = size / orig_dim1_flatten
                splited_shape = [rows]
                if len(orig_shape) >= 2:
                    splited_shape.extend(orig_shape[1:])
T
typhoonzero 已提交
872
                new_var_name = ""
873
                if self.sync_mode and add_trainer_suffix:
T
typhoonzero 已提交
874 875 876 877 878
                    new_var_name = "%s.block%d.trainer_%d" % \
                        (varname, i, self.trainer_id)
                else:
                    new_var_name = "%s.block%d" % \
                        (varname, i)
T
typhoonzero 已提交
879
                var = program.global_block().create_var(
T
typhoonzero 已提交
880 881
                    name=new_var_name,
                    persistable=False,
T
typhoonzero 已提交
882
                    dtype=orig_var.dtype,
883
                    type=orig_var.type,
T
typhoonzero 已提交
884
                    shape=splited_shape)  # flattend splited var
T
typhoonzero 已提交
885
                var_mapping[varname].append(var)
T
typhoonzero 已提交
886
            program.global_block().sync_with_cpp()
T
typhoonzero 已提交
887
        return var_mapping
T
done  
typhoonzero 已提交
888

889 890 891 892 893 894 895 896 897 898 899
    def create_splited_vars(self, source_var, block, tag):
        return [
            block.create_var(
                name=str(source_var.name + tag + str(index)),
                type=source_var.type,
                shape=source_var.shape,
                dtype=source_var.dtype)
            for index in range(len(self.pserver_endpoints))
        ]

    def _clone_var(self, block, var, persistable=True):
T
done  
typhoonzero 已提交
900 901 902 903 904 905 906
        assert isinstance(var, Variable)
        return block.create_var(
            name=var.name,
            shape=var.shape,
            dtype=var.dtype,
            type=var.type,
            lod_level=var.lod_level,
907
            persistable=persistable)
T
done  
typhoonzero 已提交
908

Y
Yancey1989 已提交
909
    def _insert_split_op(self, program, orig_var, index, splited_vars):
Y
update  
Yancey1989 已提交
910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
        if orig_var.type == core.VarDesc.VarType.SELECTED_ROWS:
            height_sections = []
            for v in splited_vars:
                height_sections.append(v.shape[0])
            program.global_block().insert_op(
                index=index + 1,
                type="split_selected_rows",
                inputs={"X": orig_var},
                outputs={"Out": splited_vars},
                attrs={"height_sections": height_sections})
        elif orig_var.type == core.VarDesc.VarType.LOD_TENSOR:
            sections = []
            for v in splited_vars:
                sections.append(v.shape[0])
            program.global_block().insert_op(
                index=index + 1,
                type="split_byref",
                inputs={"X": orig_var},
                outputs={"Out": splited_vars},
                attrs={"sections": sections}  # assume split evenly
            )
        else:
            AssertionError("Variable type should be in set "
                           "[LOD_TENSOR, SELECTED_ROWS]")
T
done  
typhoonzero 已提交
934

T
typhoonzero 已提交
935 936 937 938
    def _get_optimizer_input_shape(self, op_type, varkey, orig_shape,
                                   param_shape):
        """
        Returns the shape for optimizer inputs that need to be reshaped when
939
        Param and Grad is split to multiple servers.
T
typhoonzero 已提交
940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
        """
        # HACK(typhoonzero): Should use functions of corresponding optimizer in
        # optimizer.py to get the shape, do not  bind this in the transpiler.
        if op_type == "adam":
            if varkey in ["Moment1", "Moment2"]:
                return param_shape
        elif op_type == "adagrad":
            if varkey == "Moment":
                return param_shape
        elif op_type == "adamax":
            if varkey in ["Moment", "InfNorm"]:
                return param_shape
        elif op_type == "momentum":
            if varkey == "Velocity":
                return param_shape
        elif op_type == "":
            if varkey == "Moment":
                return param_shape
        elif op_type == "sgd":
            pass
        return orig_shape

962 963
    def _get_varname_parts(self, varname):
        # returns origin, blockid, trainerid
T
typhoonzero 已提交
964
        orig_var_name = ""
965 966 967 968 969 970 971 972 973 974
        trainer_part = ""
        block_part = ""
        trainer_idx = varname.find(".trainer_")
        if trainer_idx >= 0:
            trainer_part = varname[trainer_idx + 1:]
        else:
            trainer_idx = len(varname)
        block_index = varname.find(".block")
        if block_index >= 0:
            block_part = varname[block_index + 1:trainer_idx]
T
typhoonzero 已提交
975
        else:
976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002
            block_index = len(varname)
        orig_var_name = varname[0:min(block_index, trainer_idx)]
        return orig_var_name, block_part, trainer_part

    def _orig_varname(self, varname):
        orig, _, _ = self._get_varname_parts(varname)
        return orig

    def _append_pserver_grad_merge_ops(self, optimize_block,
                                       grad_varname_for_block, endpoint,
                                       grad_to_block_id, origin_program):
        program = optimize_block.program
        pserver_block = program.global_block()
        grad_block = None
        for g in self.param_grad_ep_mapping[endpoint]["grads"]:
            if self._orig_varname(g.name) == \
                    self._orig_varname(grad_varname_for_block):
                grad_block = g
                break
        if not grad_block:
            # do not append this op if current endpoint
            # is not dealing with this grad block
            return
        orig_varname, block_name, trainer_name = self._get_varname_parts(
            grad_block.name)
        if block_name:
            merged_var_name = '.'.join([orig_varname, block_name])
T
typhoonzero 已提交
1003
        else:
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
            merged_var_name = orig_varname
        merged_var = \
            pserver_block.vars[merged_var_name]
        grad_to_block_id.append(merged_var.name + ":" + str(optimize_block.idx))
        if self.sync_mode and self.trainer_num > 1:
            vars2merge = []
            for i in xrange(self.trainer_num):
                per_trainer_name = "%s.trainer_%d" % \
                (merged_var_name, i)
                vars2merge.append(pserver_block.vars[per_trainer_name])

            optimize_block.append_op(
                type="sum",
                inputs={"X": vars2merge},
                outputs={"Out": merged_var})
            # TODO(panyx0718): What if it's SELECTED_ROWS.
            if not merged_var.type == core.VarDesc.VarType.SELECTED_ROWS:
                optimize_block.append_op(
                    type="scale",
                    inputs={"X": merged_var},
                    outputs={"Out": merged_var},
                    attrs={"scale": 1.0 / float(self.trainer_num)})
        return merged_var
T
typhoonzero 已提交
1027

1028
    def _append_pserver_ops(self, optimize_block, opt_op, endpoint,
1029
                            grad_to_block_id, origin_program, merged_var):
1030
        program = optimize_block.program
T
typhoonzero 已提交
1031
        pserver_block = program.global_block()
T
typhoonzero 已提交
1032
        new_inputs = dict()
T
typhoonzero 已提交
1033 1034
        # update param/grad shape first, then other inputs like
        # moment can use the updated shape
T
typhoonzero 已提交
1035
        for key in opt_op.input_names:
T
typhoonzero 已提交
1036 1037 1038 1039 1040 1041
            if key == "Grad":
                new_inputs[key] = merged_var
            elif key == "Param":
                # param is already created on global program
                param_block = None
                for p in self.param_grad_ep_mapping[endpoint]["params"]:
T
typhoonzero 已提交
1042
                    if same_or_split_var(p.name, opt_op.input(key)[0]):
T
typhoonzero 已提交
1043 1044 1045 1046
                        param_block = p
                        break
                if not param_block:
                    return
T
typhoonzero 已提交
1047
                tmpvar = pserver_block.create_var(
T
typhoonzero 已提交
1048
                    name=param_block.name,
T
typhoonzero 已提交
1049
                    persistable=True,
T
typhoonzero 已提交
1050 1051 1052
                    dtype=param_block.dtype,
                    shape=param_block.shape)
                new_inputs[key] = tmpvar
1053
            elif key == "LearningRate":
1054
                # learning rate variable has already be created by non-optimize op,
1055
                # don't create it once again.
1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
                lr_varname = opt_op.input(key)[0]
                if pserver_block.vars.has_key(lr_varname):
                    new_inputs[key] = pserver_block.vars[opt_op.input(key)[0]]
                else:
                    origin_var = origin_program.global_block().vars[lr_varname]
                    tmpvar = pserver_block.create_var(
                        name=origin_var.name,
                        persistable=origin_var.persistable,
                        dtype=origin_var.dtype,
                        shape=origin_var.shape)
                    new_inputs[key] = tmpvar
T
typhoonzero 已提交
1067

T
typhoonzero 已提交
1068
        for key in opt_op.input_names:
1069 1070
            new_shape = None
            if key in ["Param", "Grad", "LearningRate"]:
T
typhoonzero 已提交
1071
                continue
1072
            var = self.origin_program.global_block().vars[opt_op.input(key)[0]]
T
typhoonzero 已提交
1073 1074 1075 1076
            # update accumulator variable shape
            param_shape = new_inputs["Param"].shape
            new_shape = self._get_optimizer_input_shape(opt_op.type, key,
                                                        var.shape, param_shape)
T
typhoonzero 已提交
1077
            tmpvar = pserver_block.create_var(
T
typhoonzero 已提交
1078 1079 1080 1081 1082
                name=var.name,
                persistable=var.persistable,
                dtype=var.dtype,
                shape=new_shape)
            new_inputs[key] = tmpvar
T
typhoonzero 已提交
1083

1084
        # change output's ParamOut variable
1085 1086
        outputs = self._get_output_map_from_op(
            self.origin_program.global_block().vars, opt_op)
1087
        outputs["ParamOut"] = new_inputs["Param"]
T
typhoonzero 已提交
1088

1089
        optimize_block.append_op(
T
typhoonzero 已提交
1090 1091
            type=opt_op.type,
            inputs=new_inputs,
T
typhoonzero 已提交
1092
            outputs=outputs,
T
typhoonzero 已提交
1093 1094
            attrs=opt_op.attrs)

1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
    def _is_splited_grad_var(self, var, var_dict):
        grad_block = None
        for _, g in var_dict.iteritems():
            if self._orig_varname(g.name) == self._orig_varname(var.name):
                if g.name.find(".trainer_") == -1:
                    grad_block = g
                    break
        return grad_block

    def _append_pserver_non_opt_ops(self, optimize_block, opt_op, endpoint):
1105
        program = optimize_block.program
1106
        # Append the ops for parameters that do not need to be optimized/updated
1107 1108
        inputs = self._get_input_map_from_op(
            self.origin_program.global_block().vars, opt_op)
1109
        for key, varlist in inputs.iteritems():
1110 1111
            if not isinstance(varlist, list):
                varlist = [varlist]
T
typhoonzero 已提交
1112
            for var in varlist:
1113 1114 1115 1116 1117 1118 1119
                # for ops like clipping and weight decay, get the splited var
                # for inputs/outputs
                grad_block = self._is_splited_grad_var(
                    var, program.global_block().vars)
                if grad_block:
                    inputs[key] = grad_block
                elif not program.global_block().vars.has_key(var.name):
1120
                    program.global_block().create_var(
T
typhoonzero 已提交
1121 1122 1123 1124 1125
                        name=var.name,
                        persistable=var.persistable,
                        dtype=var.dtype,
                        shape=var.shape)

1126 1127
        outputs = self._get_output_map_from_op(
            self.origin_program.global_block().vars, opt_op)
1128
        for key, varlist in outputs.iteritems():
1129 1130 1131
            if not isinstance(varlist, list):
                varlist = [varlist]
            for var in varlist:
1132 1133 1134 1135 1136 1137
                grad_block = self._is_splited_grad_var(
                    var, program.global_block().vars)
                if grad_block:
                    outputs[key] = grad_block
                elif not program.global_block().vars.has_key(var.name):
                    program.global_block().clone_variable(var)
1138

1139
        optimize_block.append_op(
T
typhoonzero 已提交
1140
            type=opt_op.type,
T
typhoonzero 已提交
1141 1142
            inputs=inputs,
            outputs=outputs,
T
typhoonzero 已提交
1143 1144
            attrs=opt_op.attrs)

1145 1146 1147 1148
    def _is_op_connected(self, op1, op2):
        # If one op's input is another op's output or
        # one op's output is another op's input, we say
        # the two operator is connected.
T
typhoonzero 已提交
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
        def _append_inname_remove_beta(varname_list):
            op_input_names = []
            for in_name in varname_list:
                # HACK: remove beta1 and beta2 to avoid let all
                # ops connected.
                if in_name.startswith("beta2_pow_acc") or \
                    in_name.startswith("beta1_pow_acc"):
                    continue
                else:
                    op_input_names.append(in_name)
            return op_input_names

        op1_input_names = _append_inname_remove_beta(op1.desc.input_arg_names())
T
typhoonzero 已提交
1162 1163
        op1_output_names = op1.desc.output_arg_names()

T
typhoonzero 已提交
1164
        op2_input_names = _append_inname_remove_beta(op2.desc.input_arg_names())
T
typhoonzero 已提交
1165
        op2_output_names = op2.desc.output_arg_names()
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182

        if set(op1_output_names) & set(op2_input_names) or \
           set(op1_input_names) & set(op2_output_names):
            return True
        return False

    def _create_ufind(self, optimize_ops):
        # Create a unit find data struct by optimize ops
        ufind = UnionFind(optimize_ops)
        for i in xrange(len(optimize_ops)):
            for j in xrange(i, len(optimize_ops)):
                op1 = optimize_ops[i]
                op2 = optimize_ops[j]
                if self._is_op_connected(op1, op2):
                    ufind.union(op1, op2)
        return ufind

1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
    def _is_opt_role_op(self, op):
        # NOTE: depend on oprole to find out whether this op is for
        # optimize
        op_maker = core.op_proto_and_checker_maker
        optimize_role = core.op_proto_and_checker_maker.OpRole.Optimize
        if op_maker.kOpRoleAttrName() in op.attrs and \
            int(op.attrs[op_maker.kOpRoleAttrName()]) == int(optimize_role):
            return True
        return False

    def _is_optimizer_op(self, op):
T
typhoonzero 已提交
1194 1195
        if "Param" in op.input_names and \
            "LearningRate" in op.input_names:
1196 1197 1198 1199 1200 1201 1202
            return True
        return False

    def _is_opt_op_on_pserver(self, endpoint, op):
        param_names = [
            p.name for p in self.param_grad_ep_mapping[endpoint]["params"]
        ]
T
typhoonzero 已提交
1203
        if op.input("Param")[0] in param_names:
1204 1205 1206
            return True
        else:
            for n in param_names:
T
typhoonzero 已提交
1207
                param = op.input("Param")[0]
T
typhoonzero 已提交
1208
                if same_or_split_var(n, param) and n != param:
1209 1210 1211
                    return True
            return False

T
typhoonzero 已提交
1212
    def _get_input_map_from_op(self, varmap, op):
1213
        """Returns a dict from op input name to the vars in varmap."""
T
typhoonzero 已提交
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
        iomap = dict()
        for key in op.input_names:
            vars = []
            for varname in op.input(key):
                vars.append(varmap[varname])
            if len(vars) == 1:
                iomap[key] = vars[0]
            else:
                iomap[key] = vars
        return iomap

    def _get_output_map_from_op(self, varmap, op):
1226
        """Returns a dict from op output name to the vars in varmap."""
T
typhoonzero 已提交
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
        iomap = dict()
        for key in op.output_names:
            vars = []
            for varname in op.output(key):
                vars.append(varmap[varname])
            if len(vars) == 1:
                iomap[key] = vars[0]
            else:
                iomap[key] = vars
        return iomap
1237 1238 1239 1240 1241 1242

    def _get_lr_ops(self):
        lr_ops = []
        # find learning rate variables by optimize op
        lr_vars = set()
        for op in self.optimize_ops:
1243
            if self._is_optimizer_op(op):
1244 1245 1246 1247
                lr_vars.add(op.input("LearningRate")[0])

        find_ops = []
        # find ops which output is lr var
1248
        block = self.origin_program.global_block()
1249 1250 1251 1252 1253
        for op in block.ops:
            if set(op.output_arg_names) & lr_vars:
                find_ops.append(op)
        # make a union find struct by the ops in default_main_program
        ufind = UnionFind(block.ops)
1254

1255 1256 1257 1258 1259
        for op1 in block.ops:
            for op2 in block.ops:
                # NOTE: we need to skip all optimize ops, since it is connected
                # with forward/backward ops and lr ops, we only need the lr ops.
                if op1 != op2 and self._is_op_connected(op1, op2) and \
1260
                    not self._is_optimizer_op(op1) and not self._is_optimizer_op(op2):
1261 1262 1263 1264 1265 1266
                    ufind.union(op1, op2)
        # find all ops which is related with lr var
        for op1 in block.ops:
            for op2 in find_ops:
                if ufind.is_connected(op1, op2):
                    lr_ops.append(op1)
1267 1268
                    # we only need to append op for once
                    break
1269
        return lr_ops
Y
Yancey1989 已提交
1270 1271

    def _get_optimize_pass(self):
1272 1273 1274 1275 1276 1277
        """
        Get optimizer operators, paramters and gradients from origin_program
        Returns:
            opt_ops (list): optimize operators.
            params_grads (dict): paramter->gradient.
        """
Y
Yancey1989 已提交
1278 1279 1280
        block = self.origin_program.global_block()
        opt_ops = []
        params_grads = []
1281
        origin_var_dict = self.origin_program.global_block().vars
Y
Yancey1989 已提交
1282
        for op in block.ops:
1283
            if self._is_opt_role_op(op):
Y
Yancey1989 已提交
1284
                opt_ops.append(op)
1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
                # HACK(wuyi): if we find grad vars from input of optimize
                # ops, we may get the output of clip op. Use syntax "@GRAD"
                # and op_role_var to get the pair.
                for input_name in op.input_arg_names:
                    if input_name.find("@GRAD") != -1 and \
                        op.attrs[RPC_OP_ROLE_ATTR_NAME]:
                        param_name = op.attrs[OP_ROLE_VAR_ATTR_NAME][0]
                        params_grads.append([
                            origin_var_dict[param_name],
                            origin_var_dict[input_name]
                        ])
1296 1297
            elif self._is_adam_connected_op(op):
                opt_ops.append(op)
Y
Yancey1989 已提交
1298 1299 1300
            else:
                pass
        return opt_ops, params_grads
1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312

    def _is_adam_connected_op(self, op):
        """
        A hack function to determinate whether the input operator
        is connected to optimize operator.
        """
        if op.type == "scale":
            for in_name in op.input_arg_names:
                if in_name.startswith("beta1_pow_acc") or \
                        in_name.startswith("beta2_pow_acc"):
                    return True
        return False