train.py

import functools
import os
import time
import numpy as np
import random
import datetime
import six
from collections import deque
from paddle.fluid import profiler
from ppdet.utils.utility import add_arguments, print_arguments
from paddle import fluid
from paddle.fluid.layers.learning_rate_scheduler import _decay_step_counter
from paddle.fluid.optimizer import ExponentialMovingAverage

from ppdet.experimental import mixed_precision_context
from ppdet.core.workspace import load_config, merge_config, create
from ppdet.data.reader import create_reader

from ppdet.utils import dist_utils
from ppdet.utils.eval_utils import parse_fetches, eval_run, eval_results
from ppdet.utils.stats import TrainingStats
from ppdet.utils.cli import ArgsParser
from ppdet.utils.check import check_gpu, check_version, check_config
import ppdet.utils.checkpoint as checkpoint

import logging

FORMAT = '%(asctime)s-%(levelname)s: %(message)s'
logging.basicConfig(level=logging.INFO, format=FORMAT)
logger = logging.getLogger(__name__)

parser = ArgsParser()
add_arg = functools.partial(add_arguments, argparser=parser)
parser.add_argument("--output_eval",         default='save_models/eval',    type=str, help="Evaluation directory, default is current directory.")
parser.add_argument('--vdl_log_dir',         default="logs/scalar",  type=str, help='VisualDL logging directory for scalar.')
parser.add_argument("--resume_checkpoint",   default=None,   type=str, help="Checkpoint path for resuming training.")
parser.add_argument("--fp16",                default=False,  action='store_true', help="Enable mixed precision training.")
parser.add_argument("--loss_scale",          default=8.,     type=float, help="Mixed precision training loss scale.")
parser.add_argument("--eval",                default=True,   action='store_true',  help="Whether to perform evaluation in train")
parser.add_argument("--use_vdl",             default=True,   type=bool, help="whether to record the data to VisualDL.")

# NOTE:args for profiler tools, used for benchmark
parser.add_argument('--is_profiler',         default=0, type=int,  help='The switch of profiler tools. (used for benchmark)')
parser.add_argument('--profiler_path',       default="save_models/detection.profiler", type=str,  help='The profiler output file path. (used for benchmark)')
args = parser.parse_args()


def main():
    env = os.environ
    args.dist = 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env
    if args.dist:
        trainer_id = int(env['PADDLE_TRAINER_ID'])
        local_seed = (99 + trainer_id)
        random.seed(local_seed)
        np.random.seed(local_seed)
    cfg = load_config(args.config)
    merge_config(args.opt)
    check_config(cfg)
    # check if set use_gpu=True in paddlepaddle cpu version
    check_gpu(cfg.use_gpu)
    # check if paddlepaddle version is satisfied
    check_version()

    save_only = getattr(cfg, 'save_prediction_only', False)
    if save_only:
        raise NotImplementedError('The config file only support prediction,'
                                  ' training stage is not implemented now')
    main_arch = cfg.architecture

    if cfg.use_gpu:
        devices_num = fluid.core.get_cuda_device_count()
    else:
        devices_num = int(os.environ.get('CPU_NUM', 1))

    if 'FLAGS_selected_gpus' in env:
        device_id = int(env['FLAGS_selected_gpus'])
    else:
        device_id = 0
    place = fluid.CUDAPlace(device_id) if cfg.use_gpu else fluid.CPUPlace()
    exe = fluid.Executor(place)

    lr_builder = create('LearningRate')
    optim_builder = create('OptimizerBuilder')

    # build program
    startup_prog = fluid.Program()
    train_prog = fluid.Program()
    with fluid.program_guard(train_prog, startup_prog):
        with fluid.unique_name.guard():
            model = create(main_arch)
            if args.fp16:
                assert (getattr(model.backbone, 'norm_type', None)
                        != 'affine_channel'), \
                    '--fp16 currently does not support affine channel, ' \
                    ' please modify backbone settings to use batch norm'

            with mixed_precision_context(args.loss_scale, args.fp16) as ctx:
                inputs_def = cfg['TrainReader']['inputs_def']
                feed_vars, train_loader = model.build_inputs(**inputs_def)
                train_fetches = model.train(feed_vars)
                loss = train_fetches['loss']
                if args.fp16:
                    loss *= ctx.get_loss_scale_var()
                lr = lr_builder()
                optimizer = optim_builder(lr)
                optimizer.minimize(loss)

                if args.fp16:
                    loss /= ctx.get_loss_scale_var()

            if 'use_ema' in cfg and cfg['use_ema']:
                global_steps = _decay_step_counter()
                ema = ExponentialMovingAverage(
                    cfg['ema_decay'], thres_steps=global_steps)
                ema.update()

    # parse train fetches
    train_keys, train_values, _ = parse_fetches(train_fetches)
    train_values.append(lr)

    if args.eval:
        eval_prog = fluid.Program()
        with fluid.program_guard(eval_prog, startup_prog):
            with fluid.unique_name.guard():
                model = create(main_arch)
                inputs_def = cfg['EvalReader']['inputs_def']
                feed_vars, eval_loader = model.build_inputs(**inputs_def)
                fetches = model.eval(feed_vars)
        eval_prog = eval_prog.clone(True)

        eval_reader = create_reader(cfg.EvalReader, devices_num=1)
        eval_loader.set_sample_list_generator(eval_reader, place)

        # parse eval fetches
        extra_keys = []
        if cfg.metric == 'COCO':
            extra_keys = ['im_info', 'im_id', 'im_shape']
        if cfg.metric == 'VOC':
            extra_keys = ['gt_bbox', 'gt_class', 'is_difficult']
        if cfg.metric == 'WIDERFACE':
            extra_keys = ['im_id', 'im_shape', 'gt_bbox']
        eval_keys, eval_values, eval_cls = parse_fetches(fetches, eval_prog, extra_keys)

    # compile program for multi-devices
    build_strategy = fluid.BuildStrategy()
    build_strategy.fuse_all_optimizer_ops = False
    # only enable sync_bn in multi GPU devices
    sync_bn = getattr(model.backbone, 'norm_type', None) == 'sync_bn'
    build_strategy.sync_batch_norm = sync_bn and devices_num > 1 \
                                     and cfg.use_gpu

    exec_strategy = fluid.ExecutionStrategy()
    # iteration number when CompiledProgram tries to drop local execution scopes.
    # Set it to be 1 to save memory usages, so that unused variables in
    # local execution scopes can be deleted after each iteration.
    exec_strategy.num_iteration_per_drop_scope = 1
    if args.dist:
        dist_utils.prepare_for_multi_process(exe, build_strategy, startup_prog,
                                             train_prog)
        exec_strategy.num_threads = 1

    exe.run(startup_prog)
    compiled_train_prog = fluid.CompiledProgram(train_prog).with_data_parallel(
        loss_name=loss.name,
        build_strategy=build_strategy,
        exec_strategy=exec_strategy)

    if args.eval:
        compiled_eval_prog = fluid.CompiledProgram(eval_prog)

    fuse_bn = getattr(model.backbone, 'norm_type', None) == 'affine_channel'

    ignore_params = cfg.finetune_exclude_pretrained_params \
        if 'finetune_exclude_pretrained_params' in cfg else []

    start_iter = 0
    if args.resume_checkpoint:
        checkpoint.load_checkpoint(exe, train_prog, args.resume_checkpoint)
        start_iter = checkpoint.global_step()
    elif cfg.pretrain_weights and fuse_bn and not ignore_params:
        checkpoint.load_and_fusebn(exe, train_prog, cfg.pretrain_weights)
    elif cfg.pretrain_weights:
        checkpoint.load_params(exe, train_prog, cfg.pretrain_weights, ignore_params=ignore_params)

    train_reader = create_reader(
        cfg.TrainReader, (cfg.max_iters - start_iter) * devices_num,
        cfg,
        devices_num=devices_num)
    train_loader.set_sample_list_generator(train_reader, place)

    # whether output bbox is normalized in model output layer
    is_bbox_normalized = False
    if hasattr(model, 'is_bbox_normalized') and \
            callable(model.is_bbox_normalized):
        is_bbox_normalized = model.is_bbox_normalized()

    # if map_type not set, use default 11point, only use in VOC eval
    map_type = cfg.map_type if 'map_type' in cfg else '11point'

    train_stats = TrainingStats(cfg.log_smooth_window, train_keys)
    train_loader.start()
    end_time = time.time()

    time_stat = deque(maxlen=cfg.log_smooth_window)
    best_box_ap_list = [0.0, 0]  # [map, iter]

    # use VisualDL to log data
    if args.use_vdl:
        assert six.PY3, "VisualDL requires Python >= 3.5"
        from visualdl import LogWriter
        vdl_writer = LogWriter(args.vdl_log_dir)
        vdl_loss_step = 0
        vdl_mAP_step = 0

    for it in range(start_iter, cfg.max_iters):
        start_time = end_time
        end_time = time.time()
        time_stat.append(end_time - start_time)
        time_cost = np.mean(time_stat)
        eta_sec = (cfg.max_iters - it) * time_cost
        eta = str(datetime.timedelta(seconds=int(eta_sec)))
        outs = exe.run(compiled_train_prog, fetch_list=train_values)
        stats = {k: np.array(v).mean() for k, v in zip(train_keys, outs[:-1])}

        # use vdl-paddle to log loss
        if args.use_vdl:
            if it % cfg.log_iter == 0:
                for loss_name, loss_value in stats.items():
                    vdl_writer.add_scalar(loss_name, loss_value, vdl_loss_step)
                vdl_loss_step += 1

        train_stats.update(stats)
        logs = train_stats.log()
        if it % cfg.log_iter == 0 and (not args.dist or trainer_id == 0):
            strs = 'iter: {}, lr: {:.6f}, {}, time: {:.3f}, eta: {}'.format(
                it, np.mean(outs[-1]), logs, time_cost, eta)
            logger.info(strs)

        # NOTE : profiler tools, used for benchmark
        if args.is_profiler and it == 5:
            profiler.start_profiler("All")
        elif args.is_profiler and it == 10:
            profiler.stop_profiler("total", args.profiler_path)
            return

        if (it > 0 and it % cfg.snapshot_iter == 0 or it == cfg.max_iters - 1) \
                and (not args.dist or trainer_id == 0):
            save_name = str(it) if it != cfg.max_iters - 1 else "model_final"
            if 'use_ema' in cfg and cfg['use_ema']:
                exe.run(ema.apply_program)
            checkpoint.save(exe, train_prog, os.path.join(cfg.save_dir, save_name))

            if args.eval:
                # evaluation
                resolution = None
                if 'Mask' in cfg.architecture:
                    resolution = model.mask_head.resolution
                results = eval_run(
                    exe,
                    compiled_eval_prog,
                    eval_loader,
                    eval_keys,
                    eval_values,
                    eval_cls,
                    cfg,
                    resolution=resolution)
                box_ap_stats = eval_results(
                    results, cfg.metric, cfg.num_classes, resolution,
                    is_bbox_normalized, args.output_eval, map_type,
                    cfg['EvalReader']['dataset'])

                # use vdl_paddle to log mAP
                if args.use_vdl:
                    vdl_writer.add_scalar("mAP", box_ap_stats[0], vdl_mAP_step)
                    vdl_mAP_step += 1

                if box_ap_stats[0] > best_box_ap_list[0]:
                    best_box_ap_list[0] = box_ap_stats[0]
                    best_box_ap_list[1] = it
                    checkpoint.save(exe, train_prog, os.path.join(cfg.save_dir, "best_model"))
                logger.info("Best test box ap: {}, in iter: {}".format(
                    best_box_ap_list[0], best_box_ap_list[1]))

            if 'use_ema' in cfg and cfg['use_ema']:
                exe.run(ema.restore_program)

    train_loader.reset()


if __name__ == '__main__':
    print_arguments(args)
    main()