train.py 13.0 KB
Newer Older
T
tianxin04 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
#   Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""ERNIE pretraining."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
C
chenxuyi 已提交
18 19
from __future__ import unicode_literals
from __future__ import absolute_import
T
tianxin04 已提交
20 21 22 23

import os
import time
import multiprocessing
C
chenxuyi 已提交
24
import logging
T
tianxin04 已提交
25

26
import numpy as np
T
tianxin04 已提交
27 28 29
import paddle.fluid as fluid

from reader.pretraining import ErnieDataReader
T
tianxin 已提交
30
from model.ernie_v1 import ErnieModel, ErnieConfig
T
tianxin04 已提交
31
from optimization import optimization
C
chenxuyi 已提交
32
from utils.args import print_arguments, check_cuda, prepare_logger
T
tianxin04 已提交
33 34 35 36
from utils.init import init_checkpoint, init_pretraining_params

from pretrain_args import parser

C
chenxuyi 已提交
37
log = logging.getLogger()
T
tianxin04 已提交
38
args = parser.parse_args()
T
format  
tianxin04 已提交
39

T
tianxin04 已提交
40 41
# yapf: enable.

T
format  
tianxin04 已提交
42

T
tianxin04 已提交
43 44 45 46
def create_model(pyreader_name, ernie_config):
    pyreader = fluid.layers.py_reader(
        capacity=70,
        shapes=[[-1, args.max_seq_len, 1], [-1, args.max_seq_len, 1],
Y
Yibing Liu 已提交
47
                [-1, args.max_seq_len, 1], [-1, args.max_seq_len, 1], [-1, 1],
T
tianxin04 已提交
48 49
                [-1, 1], [-1, 1]],
        dtypes=[
Y
Yibing Liu 已提交
50
            'int64', 'int64', 'int64', 'float32', 'int64', 'int64', 'int64'
T
tianxin04 已提交
51
        ],
Y
Yibing Liu 已提交
52
        lod_levels=[0, 0, 0, 0, 0, 0, 0],
T
tianxin04 已提交
53 54 55
        name=pyreader_name,
        use_double_buffer=True)

Y
Yibing Liu 已提交
56 57
    (src_ids, pos_ids, sent_ids, input_mask, mask_label, mask_pos,
     labels) = fluid.layers.read_file(pyreader)
T
tianxin04 已提交
58 59 60 61 62

    ernie = ErnieModel(
        src_ids=src_ids,
        position_ids=pos_ids,
        sentence_ids=sent_ids,
Y
Yibing Liu 已提交
63
        input_mask=input_mask,
T
tianxin04 已提交
64 65 66 67 68
        config=ernie_config,
        weight_sharing=args.weight_sharing,
        use_fp16=args.use_fp16)

    next_sent_acc, mask_lm_loss, total_loss = ernie.get_pretraining_output(
Y
Yibing Liu 已提交
69
        mask_label, mask_pos, labels)
T
tianxin04 已提交
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116

    return pyreader, next_sent_acc, mask_lm_loss, total_loss


def predict_wrapper(args,
                    exe,
                    ernie_config,
                    test_prog=None,
                    pyreader=None,
                    fetch_list=None):
    # Context to do validation.
    filelist = args.test_filelist if args.do_test else args.valid_filelist
    data_reader = ErnieDataReader(
        filelist,
        vocab_path=args.vocab_path,
        batch_size=args.batch_size,
        voc_size=ernie_config['vocab_size'],
        shuffle_files=False,
        epoch=1,
        max_seq_len=args.max_seq_len,
        is_test=True)

    if args.do_test:
        assert args.init_checkpoint is not None, "[FATAL] Please use --init_checkpoint '/path/to/checkpoints' \
                                                  to specify you pretrained model checkpoints"

        init_pretraining_params(exe, args.init_checkpoint, test_prog)

    def predict(exe=exe, pyreader=pyreader):

        pyreader.decorate_tensor_provider(data_reader.data_generator())
        pyreader.start()

        cost = 0
        lm_cost = 0
        acc = 0
        steps = 0
        time_begin = time.time()
        while True:
            try:
                each_next_acc, each_mask_lm_cost, each_total_cost = exe.run(
                    fetch_list=fetch_list, program=test_prog)
                acc += each_next_acc
                lm_cost += each_mask_lm_cost
                cost += each_total_cost
                steps += 1
                if args.do_test and steps % args.skip_steps == 0:
C
chenxuyi 已提交
117
                    log.info("[test_set] steps: %d" % steps)
T
tianxin04 已提交
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153

            except fluid.core.EOFException:
                pyreader.reset()
                break

        used_time = time.time() - time_begin
        return cost, lm_cost, acc, steps, (args.skip_steps / used_time)

    return predict


def test(args):
    ernie_config = ErnieConfig(args.ernie_config_path)
    ernie_config.print_config()

    test_prog = fluid.Program()
    test_startup = fluid.Program()
    with fluid.program_guard(test_prog, test_startup):
        with fluid.unique_name.guard():
            test_pyreader, next_sent_acc, mask_lm_loss, total_loss = create_model(
                pyreader_name='test_reader', ernie_config=ernie_config)

    test_prog = test_prog.clone(for_test=True)

    place = fluid.CUDAPlace(0) if args.use_cuda == True else fluid.CPUPlace()
    exe = fluid.Executor(place)
    exe.run(test_startup)

    predict = predict_wrapper(
        args,
        exe,
        ernie_config,
        test_prog=test_prog,
        pyreader=test_pyreader,
        fetch_list=[next_sent_acc.name, mask_lm_loss.name, total_loss.name])

C
chenxuyi 已提交
154
    log.info("test begin")
T
tianxin04 已提交
155
    loss, lm_loss, acc, steps, speed = predict()
C
chenxuyi 已提交
156
    log.info(
T
tianxin04 已提交
157 158 159 160 161 162 163
        "[test_set] loss: %f, global ppl: %f, next_sent_acc: %f, speed: %f steps/s"
        % (np.mean(np.array(loss) / steps),
           np.exp(np.mean(np.array(lm_loss) / steps)),
           np.mean(np.array(acc) / steps), speed))


def train(args):
C
chenxuyi 已提交
164
    log.info("pretraining start")
T
tianxin04 已提交
165 166 167 168 169 170 171 172 173
    ernie_config = ErnieConfig(args.ernie_config_path)
    ernie_config.print_config()

    train_program = fluid.Program()
    startup_prog = fluid.Program()
    with fluid.program_guard(train_program, startup_prog):
        with fluid.unique_name.guard():
            train_pyreader, next_sent_acc, mask_lm_loss, total_loss = create_model(
                pyreader_name='train_reader', ernie_config=ernie_config)
C
chenxuyi 已提交
174
            scheduled_lr, _ = optimization(
T
tianxin04 已提交
175 176 177 178 179 180 181 182
                loss=total_loss,
                warmup_steps=args.warmup_steps,
                num_train_steps=args.num_train_steps,
                learning_rate=args.learning_rate,
                train_program=train_program,
                startup_prog=startup_prog,
                weight_decay=args.weight_decay,
                scheduler=args.lr_scheduler,
C
chenxuyi 已提交
183 184 185 186 187 188 189 190
                use_fp16=args.use_fp16,
                use_dynamic_loss_scaling=args.use_dynamic_loss_scaling,
                init_loss_scaling=args.init_loss_scaling,
                incr_every_n_steps=args.incr_every_n_steps,
                decr_every_n_nan_or_inf=args.decr_every_n_nan_or_inf,
                incr_ratio=args.incr_ratio,
                decr_ratio=args.decr_ratio)

T
tianxin04 已提交
191 192 193 194 195 196 197 198 199

    test_prog = fluid.Program()
    with fluid.program_guard(test_prog, startup_prog):
        with fluid.unique_name.guard():
            test_pyreader, next_sent_acc, mask_lm_loss, total_loss = create_model(
                pyreader_name='test_reader', ernie_config=ernie_config)

    test_prog = test_prog.clone(for_test=True)

C
chenxuyi 已提交
200 201 202
    if len(fluid.cuda_places()) == 0:
        raise RuntimeError('not cuda device cound, check ur env setting')

T
tianxin04 已提交
203
    if args.use_cuda:
C
chenxuyi 已提交
204
        place = fluid.cuda_places()[0]
T
tianxin04 已提交
205 206 207 208 209
        dev_count = fluid.core.get_cuda_device_count()
    else:
        place = fluid.CPUPlace()
        dev_count = int(os.environ.get('CPU_NUM', multiprocessing.cpu_count()))

C
chenxuyi 已提交
210 211 212
    log.info("Device count %d" % dev_count)
    log.info("theoretical memory usage: ")
    log.info(fluid.contrib.memory_usage(
T
tianxin04 已提交
213 214 215 216
        program=train_program, batch_size=args.batch_size // args.max_seq_len))

    nccl2_num_trainers = 1
    nccl2_trainer_id = 0
C
chenxuyi 已提交
217
    log.info("args.is_distributed: %s" % args.is_distributed)
T
tianxin04 已提交
218
    if args.is_distributed:
C
chenxuyi 已提交
219
        worker_endpoints_env = os.getenv("PADDLE_TRAINER_ENDPOINTS")
T
tianxin04 已提交
220 221
        worker_endpoints = worker_endpoints_env.split(",")
        trainers_num = len(worker_endpoints)
C
chenxuyi 已提交
222
        current_endpoint = os.getenv("PADDLE_CURRENT_ENDPOINT")
T
tianxin04 已提交
223 224
        trainer_id = worker_endpoints.index(current_endpoint)
        if trainer_id == 0:
C
chenxuyi 已提交
225
            log.info("train_id == 0, sleep 60s")
T
tianxin04 已提交
226
            time.sleep(60)
C
chenxuyi 已提交
227
        log.info("worker_endpoints:{} trainers_num:{} current_endpoint:{} \
T
format  
tianxin04 已提交
228
              trainer_id:{}".format(worker_endpoints, trainers_num,
T
tianxin04 已提交
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
                                    current_endpoint, trainer_id))

        # prepare nccl2 env.
        config = fluid.DistributeTranspilerConfig()
        config.mode = "nccl2"
        t = fluid.DistributeTranspiler(config=config)
        t.transpile(
            trainer_id,
            trainers=worker_endpoints_env,
            current_endpoint=current_endpoint,
            program=train_program,
            startup_program=startup_prog)
        nccl2_num_trainers = trainers_num
        nccl2_trainer_id = trainer_id

    exe = fluid.Executor(place)
    exe.run(startup_prog)

    if args.init_checkpoint and args.init_checkpoint != "":
        init_checkpoint(exe, args.init_checkpoint, train_program, args.use_fp16)

    data_reader = ErnieDataReader(
        filelist=args.train_filelist,
        batch_size=args.batch_size,
        vocab_path=args.vocab_path,
        voc_size=ernie_config['vocab_size'],
        epoch=args.epoch,
        max_seq_len=args.max_seq_len,
        generate_neg_sample=args.generate_neg_sample)

    exec_strategy = fluid.ExecutionStrategy()
    if args.use_fast_executor:
        exec_strategy.use_experimental_executor = True
    exec_strategy.num_threads = dev_count
    exec_strategy.num_iteration_per_drop_scope = min(10, args.skip_steps)

    build_strategy = fluid.BuildStrategy()
    build_strategy.remove_unnecessary_lock = False

    train_exe = fluid.ParallelExecutor(
        use_cuda=args.use_cuda,
        loss_name=total_loss.name,
        build_strategy=build_strategy,
        exec_strategy=exec_strategy,
        main_program=train_program,
        num_trainers=nccl2_num_trainers,
        trainer_id=nccl2_trainer_id)

    if args.valid_filelist and args.valid_filelist != "":
        predict = predict_wrapper(
            args,
            exe,
            ernie_config,
            test_prog=test_prog,
            pyreader=test_pyreader,
            fetch_list=[
                next_sent_acc.name, mask_lm_loss.name, total_loss.name
            ])

    train_pyreader.decorate_tensor_provider(data_reader.data_generator())
    train_pyreader.start()
    steps = 0
    cost = []
    lm_cost = []
    acc = []
    time_begin = time.time()
    while steps < args.num_train_steps:
        try:
            steps += nccl2_num_trainers
            skip_steps = args.skip_steps * nccl2_num_trainers

            if nccl2_trainer_id != 0:
                train_exe.run(fetch_list=[])
                continue

            if steps % skip_steps != 0:
                train_exe.run(fetch_list=[])
            else:
                each_next_acc, each_mask_lm_cost, each_total_cost, np_lr = train_exe.run(
                    fetch_list=[
                        next_sent_acc.name, mask_lm_loss.name, total_loss.name,
                        scheduled_lr.name
                    ])
                acc.extend(each_next_acc)
                lm_cost.extend(each_mask_lm_cost)
                cost.extend(each_total_cost)

C
chenxuyi 已提交
316
                log.info("feed_queue size %d" % train_pyreader.queue.size())
T
tianxin04 已提交
317 318 319 320
                time_end = time.time()
                used_time = time_end - time_begin
                epoch, current_file_index, total_file, current_file, mask_type = data_reader.get_progress(
                )
C
chenxuyi 已提交
321 322
                log.info("current learning_rate:%f" % np_lr[0])
                log.info(
T
format  
tianxin04 已提交
323 324 325 326 327 328 329
                    "epoch: %d, progress: %d/%d, step: %d, loss: %f, "
                    "ppl: %f, next_sent_acc: %f, speed: %f steps/s, file: %s, mask_type: %s"
                    % (epoch, current_file_index, total_file, steps,
                       np.mean(np.array(cost)),
                       np.mean(np.exp(np.array(lm_cost))),
                       np.mean(np.array(acc)), skip_steps / used_time,
                       current_file, mask_type))
T
tianxin04 已提交
330 331 332 333 334 335 336 337 338 339 340 341
                cost = []
                lm_cost = []
                acc = []
                time_begin = time.time()

            if steps % args.save_steps == 0:
                save_path = os.path.join(args.checkpoints, "step_" + str(steps))
                fluid.io.save_persistables(exe, save_path, train_program)

            if args.valid_filelist and steps % args.validation_steps == 0:
                vali_cost, vali_lm_cost, vali_acc, vali_steps, vali_speed = predict(
                )
C
chenxuyi 已提交
342
                log.info("[validation_set] epoch: %d, step: %d, "
T
tianxin04 已提交
343 344
                      "loss: %f, global ppl: %f, batch-averged ppl: %f, "
                      "next_sent_acc: %f, speed: %f steps/s" %
T
format  
tianxin04 已提交
345
                      (epoch, steps, np.mean(np.array(vali_cost) / vali_steps),
T
tianxin04 已提交
346 347 348 349 350 351 352 353 354 355
                       np.exp(np.mean(np.array(vali_lm_cost) / vali_steps)),
                       np.mean(np.exp(np.array(vali_lm_cost) / vali_steps)),
                       np.mean(np.array(vali_acc) / vali_steps), vali_speed))

        except fluid.core.EOFException:
            train_pyreader.reset()
            break


if __name__ == '__main__':
C
chenxuyi 已提交
356
    prepare_logger(log)
T
tianxin04 已提交
357
    print_arguments(args)
T
tianxin 已提交
358
    check_cuda(args.use_cuda)
T
tianxin04 已提交
359 360 361 362
    if args.do_test:
        test(args)
    else:
        train(args)