def download_mnist()

in datasets/datasets.py [0:0]

• 33 lines of code
• 1 McCabe index (conditional complexity)

    def download_mnist(self):
        """Skips download if cache is available"""
        train_set = torchvision.datasets.MNIST(
            "/tmp/",
            train=True,
            download=True,
            transform=transforms.Compose(
                [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]
            ),
        )

        test_set = torchvision.datasets.MNIST(
            "/tmp/",
            train=False,
            download=True,
            transform=transforms.Compose(
                [transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]
            ),
        )

        (
            train_data,
            train_targets,
            valid_data,
            valid_targets,
        ) = ProjectiveMNIST.split_train_valid(train_set)
        # stratified samples
        train_data = self.sample_single_digit(
            train_data, train_targets, self.train_set_proportion
        )
        valid_data = self.sample_single_digit(
            valid_data, valid_targets, self.valid_set_proportion
        )
        test_data = self.sample_single_digit(
            test_set.data, test_set.targets, self.test_set_proportion
        )
        return train_data, valid_data, test_data