Pythonでディープラーニング入門 サンプルコード

add_save_model.py

    # Run the training
    trainer.run()

    # ここから書き足す

    # Save the trained model
    chainer.serializers.save_npz("trained_mnist.model", model)

if __name__ == '__main__':
    main()

predict_mnist.py

import chainer
import chainer.functions as F
import chainer.links as L
from PIL import Image
import numpy as np
import matplotlib.pyplot as plt

class MLP(chainer.Chain):

    def __init__(self, n_units, n_out):
        super(MLP, self).__init__()
        with self.init_scope():
            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 __call__(self, x):
        h1 = F.relu(self.l1(x))
        h2 = F.relu(self.l2(h1))
        return self.l3(h2)

model = L.Classifier(MLP(1000, 10))
chainer.serializers.load_npz('trained_mnist.model', model)

image = Image.open("number.png").convert('L')
plt.imshow(image, cmap='gray')
plt.title('input data')
plt.show()

image = np.asarray(image).astype(np.float32) / 255
image = image.reshape((1, -1))
result = model.predictor(chainer.Variable(image))

print('predicted', ':', np.argmax(result.data))
for i in range(10):
    print (str(i) , ":" , str(result.data[0,i]))

predict_mnist.sh

(base) C:\Users\Secollege160405\Desktop\mnist>python predict_mnist.py
predicted : 3
0 : -17.519121
1 : -6.611537
2 : -3.5029051
3 : 28.767323
4 : -16.44517
5 : 0.56008315
6 : -28.957706
7 : -5.5690756
8 : -3.870232
9 : -4.1418304

(base) C:\Users\Secollege160405\Desktop\mnist>

result.sh

epoch       main/loss   validation/main/loss  main/accuracy  validation/main/accuracy elapsed_time
1           0.189386    0.088241              0.943733       0.9717                   188.366
2           0.0725646   0.0875341             0.977683       0.9714                   378.107
3           0.049227    0.0849267             0.984167       0.9735                   564.709

train_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

# 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)

# さっきのコマンドでつけたオプション -g -e などの内容

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=-1, 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=1000, help='Number of units')
    parser.add_argument('--noplot', dest='plot', action='store_false', help='Disable PlotReport extension')
    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('')

    # Model の生成

    # 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.
    model = L.Classifier(MLP(args.unit, 10)) # 10 は最終的にアウトプットする値の個数
    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.Adam()
    optimizer.setup(model)

    # 以降で Iterator を生成

    # 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)

    # Trainer の生成

    # Set up a trainer
    updater = training.updaters.StandardUpdater(train_iter, optimizer, device=args.gpu)
    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))

    # 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()

train_mnist.sh

> python train_mnist.py -g -1 -e 3
GPU: -1
# unit: 1000
# Minibatch-size: 100
# epoch: 3

C:\ProgramData\Anaconda3\lib\site-packages\chainer\optimizers\adam.py:111: Runti
meWarning: invalid value encountered in sqrt
  param.data -= hp.eta * (self.lr * m / (numpy.sqrt(vhat) + hp.eps) +
epoch       main/loss   validation/main/loss  main/accuracy  validation/main/acc
uracy  elapsed_time
C:\ProgramData\Anaconda3\lib\site-packages\chainer\optimizers\adam.py:111: Runti
meWarning: invalid value encountered in sqrt
  param.data -= hp.eta * (self.lr * m / (numpy.sqrt(vhat) + hp.eps) +
1           0.189386    0.088241              0.943733       0.9717       188.366
2           0.0725646   0.0875341             0.977683       0.9714       378.107
3           0.049227    0.0849267             0.984167       0.9735       564.709