本地加载cifar10数据集pytorch

之前一直都不懂如何本地加载cifar10的数据集，联网下载太慢了

可将数据集放在当前项目的“F:/PU_prior/data/CIFAR10/” 目录下

# 本地加载CIFAR10数据集
def get_cifar10(datapath):
    train_filenames = ['F:/PU_prior/data/CIFAR10/data_batch_{}'.format(ii + 1) for ii in range(5)]
    eval_filename = 'F:/PU_prior/data/CIFAR10/test_batch'

    x_tr = np.zeros((50000, 32, 32, 3), dtype='uint8')
    y_tr = np.zeros(50000, dtype='int32')

    for ii, fname in enumerate(train_filenames):
        cur_images, cur_labels = _load_datafile(os.path.join(datapath, fname))
        x_tr[ii * 10000: (ii + 1) * 10000, ...] = cur_images
        y_tr[ii * 10000: (ii + 1) * 10000, ...] = cur_labels

    x_te, y_te = _load_datafile(os.path.join(datapath, eval_filename))
    return (x_tr, y_tr), (x_te, y_te)


(X_train, Y_train), (X_test, Y_test) = get_cifar10('')

即可得到训练集与测试集