mnist.py

#!/usr/bin/env python
import argparse

import chainer
import chainer.functions as F
import chainer.links as L
from chainer import training
from chainer.training import extensions
import dropback


# Network definition
class MLP(chainer.Chain):

    def __init__(self, n_units, n_out):
        super(MLP, self).__init__()
        with self.init_scope():
            # the size of the inputs to each layer will be inferred
            self.l1 = L.Linear(None, n_units)  # n_in -> n_units
            self.l2 = L.Linear(None, n_units)  # n_units -> n_units
            self.l3 = L.Linear(None, n_out)  # n_units -> n_out

    def forward(self, x):
        h1 = F.relu(self.l1(x))
        h2 = F.relu(self.l2(h1))
        return self.l3(h2)


def main():
    parser = argparse.ArgumentParser(description='Chainer example: MNIST')
    parser.add_argument('--batchsize', '-b', type=int, default=100,
                        help='Number of images in each mini-batch')
    parser.add_argument('--epoch', '-e', type=int, default=20,
                        help='Number of sweeps over the dataset to train')
    parser.add_argument('--frequency', '-f', type=int, default=-1,
                        help='Frequency of taking a snapshot')
    parser.add_argument('--gpu', '-g', type=int, default=0,
                        help='GPU ID (negative value indicates CPU)')
    parser.add_argument('--out', '-o', default='result',
                        help='Directory to output the result')
    parser.add_argument('--resume', '-r', default='',
                        help='Resume the training from snapshot')
    parser.add_argument('--unit', '-u', type=int, default=100,
                        help='Number of units')
    parser.add_argument('--noplot', dest='plot', action='store_false',
                        help='Disable PlotReport extension')
    parser.add_argument('-t', '--tracked_size', type=int, default=0,
                        help='Disable PlotReport extension')
    parser.add_argument('--stats', default=False, action='store_true')
    args = parser.parse_args()

    print('GPU: {}'.format(args.gpu))
    print('# unit: {}'.format(args.unit))
    print('# Minibatch-size: {}'.format(args.batchsize))
    print('# epoch: {}'.format(args.epoch))
    print('')

    # Load the MNIST dataset
    train, test = chainer.datasets.get_mnist()
    train_iter = chainer.iterators.SerialIterator(train, args.batchsize)
    test_iter = chainer.iterators.SerialIterator(test, args.batchsize,
                                                 repeat=False, shuffle=False)

    # Set up a neural network to train
    # Classifier reports softmax cross entropy loss and accuracy at every
    # iteration, which will be used by the PrintReport extension below.
    mlp = MLP(args.unit, 10)
    mlp(chainer.dataset.concat_examples(next(train_iter), -1)[0])
    model = L.Classifier(mlp)
    if args.gpu >= 0:
        # Make a specified GPU current
        chainer.backends.cuda.get_device_from_id(args.gpu).use()
        model.to_gpu()  # Copy the model to the GPU

    # Setup an optimizer
    optimizer = chainer.optimizers.SGD(lr=0.4)
    optimizer.setup(model)

    # Set up a trainer

    updater = dropback.DropBack(
        train_iter, optimizer, output_dir=args.out, device=args.gpu, tracked_size=args.tracked_size)
    trainer = training.Trainer(updater, (args.epoch, 'epoch'), out=args.out)

    # Evaluate the model with the test dataset for each epoch
    trainer.extend(extensions.Evaluator(test_iter, model, device=args.gpu))
    #decay every 25 epochs
    # Reduce the learning rate by half every 25 epochs.
    trainer.extend(extensions.ExponentialShift('lr', 0.5),
                   trigger=(25, 'epoch'))
    # 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()



if __name__ == '__main__':
    main()