Method: Chainer::Datasets::CIFAR.get_cifar

Defined in:

lib/chainer/datasets/cifar.rb

.get_cifar(n_classes, with_label, ndim, scale, device: Chainer::Device.default) ⇒ Object

# File 'lib/chainer/datasets/cifar.rb', line 14

def self.get_cifar(n_classes, with_label, ndim, scale, device: Chainer::Device.default)
  train_table = ::Datasets::CIFAR.new(n_classes: n_classes, type: :train).to_table
  test_table = ::Datasets::CIFAR.new(n_classes: n_classes, type: :test).to_table

  train_data = train_table[:pixels]
  test_data = test_table[:pixels]
  if n_classes == 10
    train_labels = train_table[:label]
    test_labels = test_table[:label]
  else
    train_labels = train_table[:fine_label]
    test_labels = test_table[:fine_label]
  end

  xm = device.xm
  [
    preprocess_cifar(xm::UInt8[*train_data], xm::UInt8[*train_labels], with_label, ndim, scale),
    preprocess_cifar(xm::UInt8[*test_data], xm::UInt8[*test_labels], with_label, ndim, scale)
  ]
end