tensorflow / datasets

TFDS is a collection of datasets ready to use with TensorFlow, Jax, ...
https://www.tensorflow.org/datasets
Apache License 2.0
4.31k stars 1.55k forks source link

Checksums for the dataset imagenette has changed #3115

Open mr-ubik opened 3 years ago

mr-ubik commented 3 years ago

Does the issue still exist with the last tfds-nightly package (pip install --upgrade tfds-nightly)?

Reproduction instructions

!rm -r ~/tensorflow_datasets/imagenette/
!rm -r ~/tensorflow_datasets/downloads
import tensorflow_datasets as tfds
tfds.load('imagenette')

Link to logs

---------------------------------------------------------------------------

NonMatchingChecksumError                  Traceback (most recent call last)

<ipython-input-25-521438880de5> in <module>()
      1 get_ipython().system('rm -r ~/tensorflow_datasets/imagenette/')
      2 get_ipython().system('rm -r ~/tensorflow_datasets/downloads')
----> 3 tfds.load('imagenette')

---

NonMatchingChecksumError: Artifact https://s3.amazonaws.com/fast-ai-imageclas/imagenette2.tgz, downloaded to /root/tensorflow_datasets/downloads/s3_fast-ai-imageclas_imagenette2s_llZjKkuLrzn8hoOF5IBPy4iCn1Rqq1SaRaYZXrNpw.tgz.tmp.137614d36be841a0bfb1615c4429f071/imagenette2.tgz,
has wrong checksum. Expected: UrlInfo(size=1.45 GiB, checksum='de32de18a7d8df0114d5e2c3c14cafb9558e8f00441a73aa7953681ad02be49c', 
filename='imagenette2.tgz'). Got: UrlInfo(size=1.45 GiB, 
checksum='6cbfac238434d89fe99e651496f0812ebc7a10fa62bd42d6874042bf01de4efd', 
filename='imagenette2.tgz').To debug, see: https://www.tensorflow.org/datasets/overview#fixing_nonmatchingchecksumerror
NikhilBartwal commented 3 years ago

@mr-ubik The imagenette dataset is successfully generated in tfds-nightly latest version. Please make sure you have installed this. (Also the expected checksums showed in your error logs are not the ones currently used in tfds-nightly) You might night to upgrade to the latest version. Thanks!

mr-ubik commented 3 years ago

@NikhilBartwal you are correct, I saw the issue persists even with tfds-nightly due to the dataset not being correctly removed in Colab.

NikhilBartwal commented 3 years ago

@mr-ubik Does the issue persists even with the latest version of tfds-nightly(pip install --upgrade tfds-nightly)? If it does, please send full error logs along with tfds.__version__

NikhilBartwal commented 3 years ago

@mr-ubik Please consider closing this issue if your query was answered. Thanks!