CIFAR-10 (python version)
=========================

A mirror, hosted for faster access from Australia. The original is at
https://www.cs.toronto.edu/~kriz/cifar.html and this file is unmodified.

  cifar-10-python.tar.gz
  md5  c58f30108f718f92721af3b95e74349a
  size 170498071 bytes

Credit: Learning Multiple Layers of Features from Tiny Images,
Alex Krizhevsky, 2009. Please cite that work rather than this mirror.

Point torchvision at it with:

  from torchvision.datasets import CIFAR10
  CIFAR10.url = "https://dl.darkovski.dev/datasets/cifar-10/cifar-10-python.tar.gz"

The checksum above is the one torchvision verifies, so its integrity
check keeps working against this mirror.