# Datalab Cup4: Unlearnable Datasets CIFAR-10 — Data

来源：https://www.kaggle.com/competitions/datalab-cup4-unlearnable-datasets-cifar-10/data

采集时间：2026-10-19T07:54:23Z

In this competition, we're classifying 10 types of objects based on their images.

# Files
This competition provides its files in Numpy array format. The samples and the corresponding labels are stored in two separate files. For instance, the label of `x_train_cifar10_unlearn[0]` is `y_train_cifar10[0]`.

- **x_train_cifar10_unlearn.npy**: training samples.
- **y_train_cifar10.npy**: training labels.
- **x_val_cifar10.npy**: pre-split training samples.
- **y_val_cifar10.npy**: pre-split training labels.
- **x_test_cifar10.npy**: test samples. You will be predicting what classes of objects these fall into.
- **sample_submission_cifar10.csv**: a sample submission file in the correct format.
    - `id`: a unique ID for each test sample. The shape of `x_test_cifar10.npy` is `(10000, 32, 32, 3)`. `x_test_cifar10[0]` represents `id = 0`.
    - `label`: predictions of images represented by the sample.