Python chainer.training.extensions.ProgressBar() Examples

def _train_trainer(self, examples):
        """Training with chainer trainer module"""
        train_iter = SerialIterator(examples, args.batch_size)
        optimizer = optimizers.Adam(alpha=args.lr)
        optimizer.setup(self.nnet)

        def loss_func(boards, target_pis, target_vs):
            out_pi, out_v = self.nnet(boards)
            l_pi = self.loss_pi(target_pis, out_pi)
            l_v = self.loss_v(target_vs, out_v)
            total_loss = l_pi + l_v
            chainer.reporter.report({
                'loss': total_loss,
                'loss_pi': l_pi,
                'loss_v': l_v,
            }, observer=self.nnet)
            return total_loss

        updater = training.StandardUpdater(
            train_iter, optimizer, device=args.device, loss_func=loss_func, converter=converter)
        # Set up the trainer.
        trainer = training.Trainer(updater, (args.epochs, 'epoch'), out=args.out)
        # trainer.extend(extensions.snapshot(), trigger=(args.epochs, 'epoch'))
        trainer.extend(extensions.LogReport())
        trainer.extend(extensions.PrintReport([
            'epoch', 'main/loss', 'main/loss_pi', 'main/loss_v', 'elapsed_time']))
        trainer.extend(extensions.ProgressBar(update_interval=10))
        trainer.run() 

def check_train(self, gpu):
        outdir = tempfile.mkdtemp()
        print("outdir: {}".format(outdir))

        n_classes = 2
        batch_size = 32

        devices = {'main': gpu}

        A = np.array([
            [0, 1, 1, 0],
            [1, 0, 0, 1],
            [1, 0, 0, 0],
            [0, 1, 0, 0],
        ]).astype(np.float32)
        model = graph_cnn.GraphCNN(A, n_out=n_classes)

        optimizer = optimizers.Adam(alpha=1e-4)
        optimizer.setup(model)
        train_dataset = EasyDataset(train=True, n_classes=n_classes)
        train_iter = chainer.iterators.MultiprocessIterator(
            train_dataset, batch_size)
        updater = ParallelUpdater(train_iter, optimizer, devices=devices)
        trainer = chainer.training.Trainer(updater, (10, 'epoch'), out=outdir)
        trainer.extend(extensions.LogReport(trigger=(1, 'epoch')))
        trainer.extend(extensions.PrintReport(
            ['epoch', 'iteration', 'main/loss', 'main/accuracy']))
        trainer.extend(extensions.ProgressBar())
        trainer.run() 

def train(args):
    nz = args.nz
    batch_size = args.batch_size
    epochs = args.epochs
    gpu = args.gpu

    # CIFAR-10 images in range [-1, 1] (tanh generator outputs)
    train, _ = datasets.get_cifar10(withlabel=False, ndim=3, scale=2)
    train -= 1.0
    train_iter = iterators.SerialIterator(train, batch_size)

    z_iter = RandomNoiseIterator(GaussianNoiseGenerator(0, 1, args.nz),
                                 batch_size)

    optimizer_generator = optimizers.RMSprop(lr=0.00005)
    optimizer_critic = optimizers.RMSprop(lr=0.00005)
    optimizer_generator.setup(Generator())
    optimizer_critic.setup(Critic())

    updater = WassersteinGANUpdater(
        iterator=train_iter,
        noise_iterator=z_iter,
        optimizer_generator=optimizer_generator,
        optimizer_critic=optimizer_critic,
        device=gpu)

    trainer = training.Trainer(updater, stop_trigger=(epochs, 'epoch'))
    trainer.extend(extensions.ProgressBar())
    trainer.extend(extensions.LogReport(trigger=(1, 'iteration')))
    trainer.extend(GeneratorSample(), trigger=(1, 'epoch'))
    trainer.extend(extensions.PrintReport(['epoch', 'iteration', 'critic/loss',
            'critic/loss/real', 'critic/loss/fake', 'generator/loss']))
    trainer.run() 

def main(model):
    train = read_data(fold=BUZZER_TRAIN_FOLD)
    valid = read_data(fold=BUZZER_DEV_FOLD)
    print('# train data: {}'.format(len(train)))
    print('# valid data: {}'.format(len(valid)))

    train_iter = chainer.iterators.SerialIterator(train, 64)
    valid_iter = chainer.iterators.SerialIterator(valid, 64, repeat=False, shuffle=False)

    optimizer = chainer.optimizers.Adam()
    optimizer.setup(model)
    optimizer.add_hook(chainer.optimizer.WeightDecay(1e-4))

    updater = training.updaters.StandardUpdater(train_iter, optimizer, converter=convert_seq, device=0)
    trainer = training.Trainer(updater, (20, 'epoch'), out=model.model_dir)

    trainer.extend(extensions.Evaluator(valid_iter, model, converter=convert_seq, device=0))

    record_trigger = training.triggers.MaxValueTrigger('validation/main/accuracy', (1, 'epoch'))
    trainer.extend(extensions.snapshot_object(model, 'buzzer.npz'), trigger=record_trigger)

    trainer.extend(extensions.LogReport())
    trainer.extend(extensions.ProgressBar())
    trainer.extend(extensions.PrintReport([
        'epoch', 'main/loss', 'validation/main/loss',
        'main/accuracy', 'validation/main/accuracy', 'elapsed_time'
    ]))

    if not os.path.isdir(model.model_dir):
        os.mkdir(model.model_dir)

    trainer.run() 

def evaluate(examples, iterator, model, converter, device,
             predict_func):
    all_results = []
    all_features = []
    total_iter = len(iterator.dataset) // iterator.batch_size + 1
    for prebatch in progressbar.ProgressBar(max_value=total_iter)(iterator):
        batch = converter(prebatch, device)
        features_list = prebatch
        # In `batch`, features is concatenated and to_gpu.
        with chainer.no_backprop_mode(), chainer.using_config('train', False):
            result = predict_func(*batch)
            for i in range(len(prebatch)):
                unique_id = int(result["unique_ids"][i])
                start_logits = [float(x) for x in result["start_logits"][i]]
                end_logits = [float(x) for x in result["end_logits"][i]]
                all_results.append(
                    RawResult(
                        unique_id=unique_id,
                        start_logits=start_logits,
                        end_logits=end_logits))
                all_features.append(features_list[i])

    output_prediction_file = os.path.join(
        FLAGS.output_dir, "predictions.json")
    output_nbest_file = os.path.join(
        FLAGS.output_dir, "nbest_predictions.json")
    write_predictions(examples, all_features, all_results,
                      FLAGS.n_best_size, FLAGS.max_answer_length,
                      FLAGS.do_lower_case, output_prediction_file,
                      output_nbest_file) 

def get_trainer(net, updater, log_dir, print_fields, curriculum=None, extra_extensions=(), epochs=10, snapshot_interval=20000, print_interval=100, postprocess=None, do_logging=True, model_files=()):
    if curriculum is None:
        trainer = chainer.training.Trainer(
            updater,
            (epochs, 'epoch'),
            out=log_dir,
        )
    else:
        trainer = chainer.training.Trainer(
            updater,
            EarlyStopIntervalTrigger(epochs, 'epoch', curriculum),
            out=log_dir,
        )

    # dump computational graph
    trainer.extend(extensions.dump_graph('main/loss'))

    # also observe learning rate
    observe_lr_extension = chainer.training.extensions.observe_lr()
    observe_lr_extension.trigger = (print_interval, 'iteration')
    trainer.extend(observe_lr_extension)

    # Take snapshots
    trainer.extend(
        extensions.snapshot(filename="trainer_snapshot"),
        trigger=lambda trainer:
        trainer.updater.is_new_epoch or
        (trainer.updater.iteration > 0 and trainer.updater.iteration % snapshot_interval == 0)
    )

    if do_logging:
        # write all statistics to a file
        trainer.extend(Logger(model_files, log_dir, keys=print_fields, trigger=(print_interval, 'iteration'), postprocess=postprocess))

        # print some interesting statistics
        trainer.extend(extensions.PrintReport(
            print_fields,
            log_report='Logger',
        ))

    # Progressbar!!
    trainer.extend(extensions.ProgressBar(update_interval=1))

    for extra_extension, trigger in extra_extensions:
        trainer.extend(extra_extension, trigger=trigger)

    return trainer 

def main(resume, gpu, load_path, data_path):
	dataset = Dataset(data_path)


	GenNetwork = MultiScaleGenerator(c.SCALE_FMS_G, c.SCALE_KERNEL_SIZES_G)
	DisNetwork = MultiScaleDiscriminator(c.SCALE_CONV_FMS_D, c.SCALE_KERNEL_SIZES_D, c.SCALE_FC_LAYER_SIZES_D)

	optimizers = {}
	optimizers["GeneratorNetwork"] = chainer.optimizers.SGD(c.LRATE_G)
	optimizers["DiscriminatorNetwork"] = chainer.optimizers.SGD(c.LRATE_D)

	iterator = chainer.iterators.SerialIterator(dataset, 1)
	params = {'LAM_ADV': 0.05, 'LAM_LP': 1, 'LAM_GDL': .1}
	updater = Updater(iterators=iterator, optimizers=optimizers,
	                  GeneratorNetwork=GenNetwork,
	                  DiscriminatorNetwork=DisNetwork,
	                  params=params,
	                  device=gpu
	                  )
	if gpu>=0:
		updater.GenNetwork.to_gpu()
		updater.DisNetwork.to_gpu()

	trainer = chainer.training.Trainer(updater, (500000, 'iteration'), out='result')
	trainer.extend(extensions.snapshot(filename='snapshot'), trigger=(1, 'iteration'))
	trainer.extend(extensions.snapshot_object(trainer.updater.GenNetwork, "GEN"))
	trainer.extend(saveGen)

	log_keys = ['epoch', 'iteration', 'GeneratorNetwork/L2Loss', 'GeneratorNetwork/GDL',
	            'DiscriminatorNetwork/DisLoss', 'GeneratorNetwork/CompositeGenLoss']
	print_keys = ['GeneratorNetwork/CompositeGenLoss','DiscriminatorNetwork/DisLoss']
	trainer.extend(extensions.LogReport(keys=log_keys, trigger=(10, 'iteration')))
	trainer.extend(extensions.PrintReport(print_keys), trigger=(10, 'iteration'))
	trainer.extend(extensions.PlotReport(['DiscriminatorNetwork/DisLoss'], 'iteration', (10, 'iteration'), file_name="DisLoss.png"))
	trainer.extend(extensions.PlotReport(['GeneratorNetwork/CompositeGenLoss'], 'iteration', (10, 'iteration'), file_name="GenLoss.png"))
	trainer.extend(extensions.PlotReport(['GeneratorNetwork/AdvLoss'], 'iteration', (10, 'iteration'), file_name="AdvGenLoss.png"))
	trainer.extend(extensions.PlotReport(['GeneratorNetwork/AdvLoss','DiscriminatorNetwork/DisLoss'], 'iteration', (10, 'iteration'), file_name="AdversarialLosses.png"))
	trainer.extend(extensions.PlotReport(['GeneratorNetwork/L2Loss'], 'iteration', (10, 'iteration'),file_name="L2Loss.png"))
	trainer.extend(extensions.PlotReport(['GeneratorNetwork/GDL'], 'iteration', (10, 'iteration'),file_name="GDL.png"))

	trainer.extend(extensions.ProgressBar(update_interval=10))
	if resume:
		# Resume from a snapshot
		chainer.serializers.load_npz(load_path, trainer)
	print(trainer.updater.__dict__)
	trainer.run() 

def run_node_classification_train(model, data,
                                  train_mask, valid_mask,
                                  epoch=10,
                                  optimizer=None,
                                  out='result',
                                  extensions_list=None,
                                  device=-1,
                                  converter=None,
                                  use_default_extensions=True,
                                  resume_path=None):
    if optimizer is None:
        # Use Adam optimizer as default
        optimizer = optimizers.Adam()
    elif not isinstance(optimizer, Optimizer):
        raise ValueError("[ERROR] optimizer must be instance of Optimizer, "
                         "but passed {}".format(type(Optimizer)))

    optimizer.setup(model)

    def one_batch_converter(batch, device):
        if not isinstance(device, Device):
            device = chainer.get_device(device)
        data, train_mask, valid_mask = batch[0]
        return (data.to_device(device),
                device.send(train_mask), device.send(valid_mask))

    data_iter = SerialIterator([(data, train_mask, valid_mask)], batch_size=1)
    updater = training.StandardUpdater(
        data_iter, optimizer, device=device,
        converter=one_batch_converter)
    trainer = training.Trainer(updater, (epoch, 'epoch'), out=out)
    if use_default_extensions:
        trainer.extend(extensions.LogReport())
        trainer.extend(AutoPrintReport())
        trainer.extend(extensions.ProgressBar(update_interval=10))
        # TODO(nakago): consider to include snapshot as default extension.
        # trainer.extend(extensions.snapshot(), trigger=(frequency, 'epoch'))

    if extensions_list is not None:
        for e in extensions_list:
            trainer.extend(e)

    if resume_path:
        chainer.serializers.load_npz(resume_path, trainer)
    trainer.run()

    return 

def train():
    parser = argparse.ArgumentParser()
    parser.add_argument('--gpu', '-g', type=int, default=-1)
    parser.add_argument('--model', '-m', type=str, default=None)
    parser.add_argument('--opt', type=str, default=None)
    parser.add_argument('--epoch', '-e', type=int, default=3)
    parser.add_argument('--lr', '-l', type=float, default=0.001)
    parser.add_argument('--inf', type=int, default=10)
    parser.add_argument('--outf', type=int, default=10)
    parser.add_argument('--batch', '-b', type=int, default=8)
    args = parser.parse_args()

    train = dataset.MovingMnistDataset(0, 7000, args.inf, args.outf)
    train_iter = iterators.SerialIterator(train, batch_size=args.batch, shuffle=True)
    test = dataset.MovingMnistDataset(7000, 10000, args.inf, args.outf)
    test_iter = iterators.SerialIterator(test, batch_size=args.batch, repeat=False, shuffle=False)

    model = network.MovingMnistNetwork(sz=[128,64,64], n=2)

    if args.model != None:
        print( "loading model from " + args.model )
        serializers.load_npz(args.model, model)

    if args.gpu >= 0:
        cuda.get_device_from_id(0).use()
        model.to_gpu()

    opt = optimizers.Adam(alpha=args.lr)
    opt.setup(model)

    if args.opt != None:
        print( "loading opt from " + args.opt )
        serializers.load_npz(args.opt, opt)

    updater = training.StandardUpdater(train_iter, opt, device=args.gpu)
    trainer = training.Trainer(updater, (args.epoch, 'epoch'), out='results')

    trainer.extend(extensions.Evaluator(test_iter, model, device=args.gpu))
    trainer.extend(extensions.LogReport(trigger=(10, 'iteration')))
    trainer.extend(extensions.PrintReport(['epoch', 'main/loss', 'validation/main/loss']))
    trainer.extend(extensions.ProgressBar(update_interval=1))

    trainer.run()

    modelname = "./results/model"
    print( "saving model to " + modelname )
    serializers.save_npz(modelname, model)

    optname = "./results/opt"
    print( "saving opt to " + optname )
    serializers.save_npz(optname, opt) 

def register_extensions(trainer, model, test_iter, args):
    if args.mode.startswith('seg'):
        # Max accuracy
        best_trigger = training.triggers.BestValueTrigger(
            'validation/main/accuracy', lambda a, b: a < b, (1, 'epoch'))
    elif args.mode.startswith('mat'):
        # Min loss
        best_trigger = training.triggers.BestValueTrigger(
            'validation/main/loss', lambda a, b: a > b, (1, 'epoch'))
    else:
        logger.error('Invalid training mode')

    # Segmentation extensions
    trainer.extend(
        custom_extensions.PortraitVisEvaluator(
            test_iter, model, device=args.gpus[0],
            converter=select_converter(args.mode),
            filename='vis_epoch={epoch}_idx={index}.jpg',
            mode=args.mode
        ), trigger=(1, 'epoch'))

    # Basic extensions
    trainer.extend(extensions.dump_graph('main/loss'))
    trainer.extend(extensions.LogReport(trigger=(200, 'iteration')))
    trainer.extend(extensions.ProgressBar(update_interval=20))
    trainer.extend(extensions.PrintReport(
        ['epoch', 'main/loss', 'validation/main/loss', 'main/accuracy',
         'validation/main/accuracy', 'lr', 'elapsed_time']))
    trainer.extend(extensions.observe_lr(), trigger=(200, 'iteration'))

    # Snapshots
    trainer.extend(extensions.snapshot(
        filename='snapshot_epoch_{.updater.epoch}'
    ), trigger=(5, 'epoch'))
    trainer.extend(extensions.snapshot_object(
        model, filename='model_best'
    ), trigger=best_trigger)

    # ChainerUI extensions
    trainer.extend(chainerui.extensions.CommandsExtension())
    chainerui.utils.save_args(args, args.out)

    # Plotting extensions
    if extensions.PlotReport.available():
        trainer.extend(
            extensions.PlotReport(
                ['main/loss', 'validation/main/loss'],
                'epoch', file_name='loss.png'))
        trainer.extend(
            extensions.PlotReport(
                ['main/accuracy', 'validation/main/accuracy'],
                'epoch', file_name='accuracy.png')) 

def main():
    gpu, out = -1, "result"
    stepsize = 0.001
    batchsize, epoch = 10000, 10
    beta, gamma = 0., 1.

    data_id, prior = 0, .5
    n_p, n_n, n_u, n_t, n_vp, n_vn, n_vu = 100, 0, 10000, 100, 20, 20, 100
    data_name, x_p, x_n, x_u, y_u, x_t, y_t, x_vp, x_vn, x_vu, y_vu \
        = load_dataset(data_id, n_p, n_n, n_u, prior, n_t, n_vp=n_vp, n_vn=n_vn, n_vu=n_vu)

    x_p, x_n, x_u, x_t, x_vp, x_vn, x_vu = x_p.astype(np.float32), x_n.astype(np.float32), \
        x_u.astype(np.float32), x_t.astype(np.float32), x_vp.astype(np.float32), \
        x_vn.astype(np.float32), x_vu.astype(np.float32), 
    XYtrain = TupleDataset(np.r_[x_p, x_u], np.r_[np.ones(100), np.zeros(10000)].astype(np.int32))
    XYtest = TupleDataset(np.r_[x_vp, x_vu], np.r_[np.ones(20), np.zeros(100)].astype(np.int32))
    train_iter = chainer.iterators.SerialIterator(XYtrain, batchsize)
    test_iter = chainer.iterators.SerialIterator(XYtest, batchsize, repeat=False, shuffle=False)

    loss_type = lambda x: F.sigmoid(-x)
    nnpu_risk = PU_Risk(prior, loss=loss_type, nnPU=True, gamma=gamma, beta=beta)
    pu_acc = PU_Accuracy(prior)

    model = L.Classifier(MLP(), lossfun=nnpu_risk, accfun=pu_acc)
    if gpu >= 0:
        chainer.backends.cuda.get_device_from_id(gpu).use()
        model.to_gpu(gpu)

    optimizer = chainer.optimizers.Adam(alpha=stepsize)
    optimizer.setup(model)
    optimizer.add_hook(chainer.optimizer.WeightDecay(0.005))

    updater = chainer.training.StandardUpdater(train_iter, optimizer, device=gpu)
    trainer = chainer.training.Trainer(updater, (epoch, 'epoch'), out=out)
    trainer.extend(extensions.LogReport(trigger=(1, 'epoch')))
    trainer.extend(extensions.Evaluator(test_iter, model, device=gpu))
    trainer.extend(extensions.ProgressBar())
    trainer.extend(extensions.PrintReport(
                ['epoch', 'main/loss', 'validation/main/loss',
                 'main/accuracy', 'validation/main/accuracy', 
                 'elapsed_time']))
    key = 'validation/main/accuracy'
    model_name = 'model'
    trainer.extend(extensions.snapshot_object(model, model_name),
                   trigger=chainer.training.triggers.MaxValueTrigger(key))
    if extensions.PlotReport.available():
            trainer.extend(
                extensions.PlotReport(['main/loss', 'validation/main/loss'], 'epoch', file_name=f'loss_curve.png'))
            trainer.extend(
                extensions.PlotReport(['main/accuracy', 'validation/main/accuracy'], 
                                      'epoch', file_name=f'accuracy_curve.png'))


    trainer.run()

    yh = pred(model, x_t, batchsize, gpu)
    mr = prior*np.mean(yh[y_t == +1] <= 0) + (1-prior)*np.mean(yh[y_t == -1] >= 0)
    print("mr: {}".format(mr)) 

def train_one_epoch(model, train_data, lr, gpu, batchsize, out):
    train_model = PixelwiseSoftmaxClassifier(model)
    if gpu >= 0:
        # Make a specified GPU current
        chainer.cuda.get_device_from_id(gpu).use()
        train_model.to_gpu()  # Copy the model to the GPU
    log_trigger = (0.1, 'epoch')
    validation_trigger = (1, 'epoch')
    end_trigger = (1, 'epoch')

    train_data = TransformDataset(
        train_data, ('img', 'label_map'), SimpleDoesItTransform(model.mean))
    val = VOCSemanticSegmentationWithBboxDataset(
        split='val').slice[:, ['img', 'label_map']]

    # Iterator
    train_iter = iterators.MultiprocessIterator(train_data, batchsize)
    val_iter = iterators.MultiprocessIterator(
        val, 1, shuffle=False, repeat=False, shared_mem=100000000)

    # Optimizer
    optimizer = optimizers.MomentumSGD(lr=lr, momentum=0.9)
    optimizer.setup(train_model)
    optimizer.add_hook(chainer.optimizer_hooks.WeightDecay(rate=0.0001))

    # Updater
    updater = training.updaters.StandardUpdater(
        train_iter, optimizer, device=gpu)

    # Trainer
    trainer = training.Trainer(updater, end_trigger, out=out)

    trainer.extend(extensions.LogReport(trigger=log_trigger))
    trainer.extend(extensions.observe_lr(), trigger=log_trigger)
    trainer.extend(extensions.dump_graph('main/loss'))

    if extensions.PlotReport.available():
        trainer.extend(extensions.PlotReport(
            ['main/loss'], x_key='iteration',
            file_name='loss.png'))
        trainer.extend(extensions.PlotReport(
            ['validation/main/miou'], x_key='iteration',
            file_name='miou.png'))

    trainer.extend(extensions.snapshot_object(
        model, filename='snapshot.npy'),
        trigger=end_trigger)
    trainer.extend(extensions.PrintReport(
        ['epoch', 'iteration', 'elapsed_time', 'lr',
         'main/loss', 'validation/main/miou',
         'validation/main/mean_class_accuracy',
         'validation/main/pixel_accuracy']),
        trigger=log_trigger)
    trainer.extend(extensions.ProgressBar(update_interval=10))

    trainer.extend(
        SemanticSegmentationEvaluator(
            val_iter, model,
            voc_semantic_segmentation_label_names),
        trigger=validation_trigger)
    trainer.run() 

def train(train_data_path, test_data_path, args):
    vocab = collections.defaultdict(lambda: len(vocab))
    vocab['<unk>'] = 0

    train_data = babi.read_data(vocab, train_data_path)
    test_data = babi.read_data(vocab, test_data_path)
    print('Training data: %s: %d' % (train_data_path, len(train_data)))
    print('Test data: %s: %d' % (test_data_path, len(test_data)))

    train_data = memnn.convert_data(train_data, args.max_memory)
    test_data = memnn.convert_data(test_data, args.max_memory)

    encoder = memnn.make_encoder(args.sentence_repr)
    network = memnn.MemNN(
        args.unit, len(vocab), encoder, args.max_memory, args.hop)
    model = chainer.links.Classifier(network, label_key='answer')
    opt = chainer.optimizers.Adam()

    if args.gpu >= 0:
        chainer.cuda.get_device(args.gpu).use()
        model.to_gpu()

    opt.setup(model)

    train_iter = chainer.iterators.SerialIterator(
        train_data, args.batchsize)
    test_iter = chainer.iterators.SerialIterator(
        test_data, args.batchsize, repeat=False, shuffle=False)
    updater = chainer.training.StandardUpdater(
        train_iter, opt, device=args.gpu)
    trainer = chainer.training.Trainer(updater, (args.epoch, 'epoch'))

    @chainer.training.make_extension()
    def fix_ignore_label(trainer):
        network.fix_ignore_label()

    trainer.extend(fix_ignore_label)
    trainer.extend(extensions.Evaluator(test_iter, model, device=args.gpu))
    trainer.extend(extensions.LogReport())
    trainer.extend(extensions.PrintReport(
        ['epoch', 'main/loss', 'validation/main/loss',
         'main/accuracy', 'validation/main/accuracy']))
    trainer.extend(extensions.ProgressBar(update_interval=10))
    trainer.run()

    if args.model:
        memnn.save_model(args.model, model, vocab) 

def get_trainer(optimizer, iter_train, iter_valid, iter_valid_raw,
                class_names, args):
    model = optimizer.target

    updater = chainer.training.StandardUpdater(
        iter_train, optimizer, device=args.gpu)

    trainer = chainer.training.Trainer(
        updater, (args.max_iteration, 'iteration'), out=args.out)

    trainer.extend(fcn.extensions.ParamsReport(args.__dict__))

    trainer.extend(extensions.ProgressBar(update_interval=5))

    trainer.extend(extensions.LogReport(
        trigger=(args.interval_print, 'iteration')))
    trainer.extend(extensions.PrintReport(
        ['epoch', 'iteration', 'elapsed_time',
         'main/loss', 'validation/main/miou']))

    def pred_func(x):
        model(x)
        return model.score

    trainer.extend(
        fcn.extensions.SemanticSegmentationVisReport(
            pred_func, iter_valid_raw,
            transform=fcn.datasets.transform_lsvrc2012_vgg16,
            class_names=class_names, device=args.gpu, shape=(4, 2)),
        trigger=(args.interval_eval, 'iteration'))

    trainer.extend(
        chainercv.extensions.SemanticSegmentationEvaluator(
            iter_valid, model, label_names=class_names),
        trigger=(args.interval_eval, 'iteration'))

    trainer.extend(extensions.snapshot_object(
        target=model, filename='model_best.npz'),
        trigger=chainer.training.triggers.MaxValueTrigger(
            key='validation/main/miou',
            trigger=(args.interval_eval, 'iteration')))

    assert extensions.PlotReport.available()
    trainer.extend(extensions.PlotReport(
        y_keys=['main/loss'], x_key='iteration',
        file_name='loss.png', trigger=(args.interval_print, 'iteration')))
    trainer.extend(extensions.PlotReport(
        y_keys=['validation/main/miou'], x_key='iteration',
        file_name='miou.png', trigger=(args.interval_print, 'iteration')))

    return trainer 

def train(train_data_path, test_data_path, args):
    device = chainer.get_device(args.device)
    device.use()

    vocab = collections.defaultdict(lambda: len(vocab))
    vocab['<unk>'] = 0

    train_data = babi.read_data(vocab, train_data_path)
    test_data = babi.read_data(vocab, test_data_path)
    print('Training data: %s: %d' % (train_data_path, len(train_data)))
    print('Test data: %s: %d' % (test_data_path, len(test_data)))

    train_data = memnn.convert_data(train_data, args.max_memory)
    test_data = memnn.convert_data(test_data, args.max_memory)

    encoder = memnn.make_encoder(args.sentence_repr)
    network = memnn.MemNN(
        args.unit, len(vocab), encoder, args.max_memory, args.hop)
    model = chainer.links.Classifier(network, label_key='answer')
    opt = chainer.optimizers.Adam()

    model.to_device(device)

    opt.setup(model)

    train_iter = chainer.iterators.SerialIterator(
        train_data, args.batchsize)
    test_iter = chainer.iterators.SerialIterator(
        test_data, args.batchsize, repeat=False, shuffle=False)
    updater = chainer.training.StandardUpdater(train_iter, opt, device=device)
    trainer = chainer.training.Trainer(updater, (args.epoch, 'epoch'))

    @chainer.training.make_extension()
    def fix_ignore_label(trainer):
        network.fix_ignore_label()

    trainer.extend(fix_ignore_label)
    trainer.extend(extensions.Evaluator(test_iter, model, device=device))
    trainer.extend(extensions.LogReport())
    trainer.extend(extensions.PrintReport(
        ['epoch', 'main/loss', 'validation/main/loss',
         'main/accuracy', 'validation/main/accuracy']))
    trainer.extend(extensions.ProgressBar(update_interval=10))
    trainer.run()

    if args.model:
        memnn.save_model(args.model, model, vocab) 

def train_main(args):
    """
    trains model specfied in args.
    main method for train subcommand.
    """
    # load text
    with open(args.text_path) as f:
        text = f.read()
    logger.info("corpus length: %s.", len(text))

    # data iterator
    data_iter = DataIterator(text, args.batch_size, args.seq_len)

    # load or build model
    if args.restore:
        logger.info("restoring model.")
        load_path = args.checkpoint_path if args.restore is True else args.restore
        model = load_model(load_path)
    else:
        net = Network(vocab_size=VOCAB_SIZE,
                      embedding_size=args.embedding_size,
                      rnn_size=args.rnn_size,
                      num_layers=args.num_layers,
                      drop_rate=args.drop_rate)
        model = L.Classifier(net)

    # make checkpoint directory
    log_dir = make_dirs(args.checkpoint_path)
    with open("{}.json".format(args.checkpoint_path), "w") as f:
        json.dump(model.predictor.args, f, indent=2)
    chainer.serializers.save_npz(args.checkpoint_path, model)
    logger.info("model saved: %s.", args.checkpoint_path)

    # optimizer
    optimizer = chainer.optimizers.Adam(alpha=args.learning_rate)
    optimizer.setup(model)
    # clip gradient norm
    optimizer.add_hook(chainer.optimizer.GradientClipping(args.clip_norm))

    # trainer
    updater = BpttUpdater(data_iter, optimizer)
    trainer = chainer.training.Trainer(updater, (args.num_epochs, 'epoch'), out=log_dir)
    trainer.extend(extensions.snapshot_object(model, filename=os.path.basename(args.checkpoint_path)))
    trainer.extend(extensions.ProgressBar(update_interval=1))
    trainer.extend(extensions.LogReport())
    trainer.extend(extensions.PlotReport(y_keys=["main/loss"]))
    trainer.extend(LoggerExtension(text))

    # training start
    model.predictor.reset_state()
    logger.info("start of training.")
    time_train = time.time()
    trainer.run()

    # training end
    duration_train = time.time() - time_train
    logger.info("end of training, duration: %ds.", duration_train)
    # generate text
    seed = generate_seed(text)
    generate_text(model, seed, 1024, 3)
    return model 

def prepare_trainer(net,
                    optimizer_name,
                    lr,
                    momentum,
                    num_epochs,
                    train_iter,
                    val_iter,
                    logging_dir_path,
                    num_gpus=0):
    if optimizer_name == "sgd":
        optimizer = chainer.optimizers.MomentumSGD(lr=lr, momentum=momentum)
    elif optimizer_name == "nag":
        optimizer = chainer.optimizers.NesterovAG(lr=lr, momentum=momentum)
    else:
        raise Exception('Unsupported optimizer: {}'.format(optimizer_name))
    optimizer.setup(net)

    # devices = tuple(range(num_gpus)) if num_gpus > 0 else (-1, )
    devices = (0,) if num_gpus > 0 else (-1,)

    updater = training.updaters.StandardUpdater(
        iterator=train_iter,
        optimizer=optimizer,
        device=devices[0])
    trainer = training.Trainer(
        updater=updater,
        stop_trigger=(num_epochs, 'epoch'),
        out=logging_dir_path)

    val_interval = 100000, 'iteration'
    log_interval = 1000, 'iteration'

    trainer.extend(
        extension=extensions.Evaluator(
            val_iter,
            net,
            device=devices[0]),
        trigger=val_interval)
    trainer.extend(extensions.dump_graph('main/loss'))
    trainer.extend(extensions.snapshot(), trigger=val_interval)
    trainer.extend(
        extensions.snapshot_object(
            net,
            'model_iter_{.updater.iteration}'),
        trigger=val_interval)
    trainer.extend(extensions.LogReport(trigger=log_interval))
    trainer.extend(extensions.observe_lr(), trigger=log_interval)
    trainer.extend(
        extensions.PrintReport([
            'epoch', 'iteration', 'main/loss', 'validation/main/loss', 'main/accuracy', 'validation/main/accuracy',
            'lr']),
        trigger=log_interval)
    trainer.extend(extensions.ProgressBar(update_interval=10))

    return trainer 

def prepare_trainer(net,
                    optimizer_name,
                    lr,
                    momentum,
                    num_epochs,
                    train_iter,
                    val_iter,
                    logging_dir_path,
                    num_gpus=0):
    if optimizer_name == "sgd":
        optimizer = chainer.optimizers.MomentumSGD(lr=lr, momentum=momentum)
    elif optimizer_name == "nag":
        optimizer = chainer.optimizers.NesterovAG(lr=lr, momentum=momentum)
    else:
        raise Exception('Unsupported optimizer: {}'.format(optimizer_name))
    optimizer.setup(net)

    # devices = tuple(range(num_gpus)) if num_gpus > 0 else (-1, )
    devices = (0,) if num_gpus > 0 else (-1,)

    updater = training.updaters.StandardUpdater(
        iterator=train_iter,
        optimizer=optimizer,
        device=devices[0])
    trainer = training.Trainer(
        updater=updater,
        stop_trigger=(num_epochs, 'epoch'),
        out=logging_dir_path)

    val_interval = 100000, 'iteration'
    log_interval = 1000, 'iteration'

    trainer.extend(
        extension=extensions.Evaluator(
            val_iter,
            net,
            device=devices[0]),
        trigger=val_interval)
    trainer.extend(extensions.dump_graph('main/loss'))
    trainer.extend(extensions.snapshot(), trigger=val_interval)
    trainer.extend(
        extensions.snapshot_object(
            net,
            'model_iter_{.updater.iteration}'),
        trigger=val_interval)
    trainer.extend(extensions.LogReport(trigger=log_interval))
    trainer.extend(extensions.observe_lr(), trigger=log_interval)
    trainer.extend(
        extensions.PrintReport([
            'epoch', 'iteration', 'main/loss', 'validation/main/loss', 'main/accuracy', 'validation/main/accuracy',
            'lr']),
        trigger=log_interval)
    trainer.extend(extensions.ProgressBar(update_interval=10))

    return trainer 

def prepare_trainer(net,
                    optimizer_name,
                    lr,
                    momentum,
                    num_epochs,
                    train_data,
                    val_data,
                    logging_dir_path,
                    use_gpus):
    if optimizer_name == "sgd":
        optimizer = chainer.optimizers.MomentumSGD(lr=lr, momentum=momentum)
    elif optimizer_name == "nag":
        optimizer = chainer.optimizers.NesterovAG(lr=lr, momentum=momentum)
    else:
        raise Exception("Unsupported optimizer: {}".format(optimizer_name))
    optimizer.setup(net)

    # devices = tuple(range(num_gpus)) if num_gpus > 0 else (-1, )
    devices = (0,) if use_gpus else (-1,)

    updater = training.updaters.StandardUpdater(
        iterator=train_data["iterator"],
        optimizer=optimizer,
        device=devices[0])
    trainer = training.Trainer(
        updater=updater,
        stop_trigger=(num_epochs, "epoch"),
        out=logging_dir_path)

    val_interval = 100000, "iteration"
    log_interval = 1000, "iteration"

    trainer.extend(
        extension=extensions.Evaluator(
            iterator=val_data["iterator"],
            target=net,
            device=devices[0]),
        trigger=val_interval)
    trainer.extend(extensions.dump_graph("main/loss"))
    trainer.extend(extensions.snapshot(), trigger=val_interval)
    trainer.extend(
        extensions.snapshot_object(
            net,
            "model_iter_{.updater.iteration}"),
        trigger=val_interval)
    trainer.extend(extensions.LogReport(trigger=log_interval))
    trainer.extend(extensions.observe_lr(), trigger=log_interval)
    trainer.extend(
        extensions.PrintReport([
            "epoch", "iteration", "main/loss", "validation/main/loss", "main/accuracy", "validation/main/accuracy",
            "lr"]),
        trigger=log_interval)
    trainer.extend(extensions.ProgressBar(update_interval=10))

    return trainer 

def run_training(args, model):
    trainer = create_trainer(args, model)

    # Dump a computational graph from 'loss' variable at the first iteration
    # The "main" refers to the target link of the "main" optimizer.
    trainer.extend(extensions.dump_graph('main/loss'))

    # Take a snapshot for each specified epoch
    frequency = args.epoch if args.frequency == -1 else max(1, args.frequency)
    trainer.extend(extensions.snapshot(), trigger=(frequency, 'epoch'))

    # Write a log of evaluation statistics for each epoch
    trainer.extend(extensions.LogReport())

    # Save two plot images to the result dir
    if args.plot and extensions.PlotReport.available():
        trainer.extend(
            extensions.PlotReport(['main/loss', 'validation/main/loss'],
                                  'epoch', file_name='loss.png'))
        trainer.extend(
            extensions.PlotReport(
                ['main/accuracy', 'validation/main/accuracy'],
                'epoch', file_name='accuracy.png'))

    # Print selected entries of the log to stdout
    # Here "main" refers to the target link of the "main" optimizer again, and
    # "validation" refers to the default name of the Evaluator extension.
    # Entries other than 'epoch' are reported by the Classifier link, called by
    # either the updater or the evaluator.
    trainer.extend(extensions.PrintReport(
        ['epoch', 'main/loss', 'validation/main/loss',
         'main/accuracy', 'validation/main/accuracy', 'elapsed_time']))

    # Print a progress bar to stdout
    trainer.extend(extensions.ProgressBar())

    if args.resume:
        # Resume from a snapshot
        chainer.serializers.load_npz(args.resume, trainer)

    # Run the training
    trainer.run() 

def run_training(args, model):
    trainer = create_trainer(args, model)

    # Dump a computational graph from 'loss' variable at the first iteration
    # The "main" refers to the target link of the "main" optimizer.
    trainer.extend(extensions.dump_graph('main/loss'))

    # Take a snapshot for each specified epoch
    frequency = args.epoch if args.frequency == -1 else max(1, args.frequency)
    trainer.extend(extensions.snapshot(), trigger=(frequency, 'epoch'))

    # Write a log of evaluation statistics for each epoch
    trainer.extend(extensions.LogReport())

    # Save two plot images to the result dir
    if args.plot and extensions.PlotReport.available():
        trainer.extend(
            extensions.PlotReport(['main/loss', 'validation/main/loss'],
                                  'epoch', file_name='loss.png'))
        trainer.extend(
            extensions.PlotReport(
                ['main/accuracy', 'validation/main/accuracy'],
                'epoch', file_name='accuracy.png'))

    # Print selected entries of the log to stdout
    # Here "main" refers to the target link of the "main" optimizer again, and
    # "validation" refers to the default name of the Evaluator extension.
    # Entries other than 'epoch' are reported by the Classifier link, called by
    # either the updater or the evaluator.
    trainer.extend(extensions.PrintReport(
        ['epoch', 'main/loss', 'validation/main/loss',
         'main/accuracy', 'validation/main/accuracy', 'elapsed_time']))

    # Print a progress bar to stdout
    trainer.extend(extensions.ProgressBar())

    if args.resume:
        # Resume from a snapshot
        chainer.serializers.load_npz(args.resume, trainer)

    # Run the training
    trainer.run()