# DLchallengeMNIST — Data

来源：https://www.kaggle.com/competitions/dlchmnist/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

## Files

*  The train set is available with demo code at:

[https://tutoria.ualg.pt/2025/mod/folder/view.php?id=116240](https://tutoria.ualg.pt/2025/mod/folder/view.php?id=116240)

*   **eval.npy**  - the evaluation set

It is stored as a numpy array and can be loaded with:

```
import numpy as np

np.load(evalX.npy)
```

This dataset has 70,000 samples, each with 400 values, sorted from the top left corner of a 20x20 digit image.

The evaluation dataset is shuffled and stores only the array X with the sample data.
There is no information on the labels.

The goal is to predict the correct label for each sample.

*   **sample_submission.csv** - a sample submission file in the correct format, with the header and the first rows