Checksums for oxford_flowers102 changed

[lgeiger]

Short description
The checksum for oxford_flowers102 changed recently which breaks downloading of the dataset.

Environment information

• Operating System: macOS
• Python version: 3.7
• tensorflow-datasets/tfds-nightly version: 3.1.0
• tensorflow/tensorflow-gpu/tf-nightly/tf-nightly-gpu version: 2.2.0rc3

Reproduction instructions

rm -r ~/tensorflow_datasets/oxford_flowers102/
rm -r ~/tensorflow_datasets/downloads

tfds.load("oxford_flowers102", split="train")

Link to logs

Artifact https://www.robots.ox.ac.uk/~vgg/data/flowers/102/102flowers.tgz, downloaded to ~/tensorflow_datasets/downloads/robots.ox.ac.uk_vgg_flowers_102_102flowersoWedSp98maBn1wypsDib6T-q2NVbO40fwvTflmPmQpY.tgz.tmp.9496e46fb0ed404fa5b339ba8c39a725/102flowers.tgz, has wrong checksum. This might indicate:
 * The website may be down (e.g. returned a 503 status code). Please check the url.
 * For Google Drive URLs, try again later as Drive sometimes rejects downloads when too many people access the same URL. See https://github.com/tensorflow/datasets/issues/1482
 * The original datasets files may have been updated. In this case the TFDS dataset builder should be updated to use the new files and checksums. Sorry about that. Please open an issue or send us a PR with a fix.
 * If you're adding a new dataset, don't forget to register the checksums as explained in: https://www.tensorflow.org/datasets/add_dataset#2_run_download_and_prepare_locally

Expected behavior
Checksums should match and extraction would succeed.

[elfelround]

probably this:

Sorry, you can't view or download this file at this time.

Too many users have viewed or downloaded this file recently. Please try accessing the file again later. If the file you are trying to access is particularly large or is shared with many people, it may take up to 24 hours to be able to view or download the file. If you still can't access a file after 24 hours, contact your domain administrator.
© 2020 Google - Help - Privacy & Terms

[lgeiger]

@elfelround Looks like the file is still available at https://www.robots.ox.ac.uk/~vgg/data/flowers/102/102flowers.tgz so something probably changed.

[elfelround]

well if ur downloads download the tgz why not unzip it outside of downloads and see if it compiles?
could be that the centralised checksum is different from the one on the site too? one of them updated or tampered, or some sort of cdn download limit

[vijayphoenix]

@lgeiger @elfelround It seems that http://www.robots.ox.ac.uk/~vgg/data/ site had some changes.

I have updated the checksums.
Thanks for reporting the issue.

[lgeiger]

Thanks for the fix 👍

[harshag24]

The problem still persists! Is it due to too many users using the file because even after @vijayphoenix has updated checksums it persists?

[Eshan-Agarwal]

Can you try with tfds-nightly ? Changes will appear in tfds-nightly soon in few days. Meanwhile you can apply changes from PR

[danielmapar]

Tried %pip install tfds-nightly on my local machine with no success. Any suggestions?

[danielmapar]

pip --no-cache-dir install tfds-nightly fixes the issue for me

[Eshan-Agarwal]

Here is colab with tfds-nightly working with no issue.

[mikechen66]

lgeiger and danielmapar are correct.

I solve the issue by removing the old files and installing tfds-nightly

1. Remove the downloaded files

rm -rf ~/tensorflow_datasets/oxford_flowers102/
rm -rf ~/tensorflow_datasets/downloads

2. Install tfds-nightly

$ pip --no-cache-dir install tfds-nightly

Thank a lot.