Datasets: [GSOC] Checksum mismatch for Pix3D

[aradhyamathur]

While implementing the dataset for Pix3D I run into error due to a checksum mismatch from the DownloadManager. The error is as follows -

`~/anaconda3/envs/gsoc/lib/python3.7/site-packages/tensorflow_datasets-2.1.0-py3.7.egg/tensorflow_datasets/core/download/download_manager.py in callback(val)
260 checksum, dl_size = val
261 return self._handle_download_result(
--> 262 resource, download_dir_path, checksum, dl_size)
263 return self._downloader.download(url, download_dir_path).then(callback)
264

~/anaconda3/envs/gsoc/lib/python3.7/site-packages/tensorflow_datasets-2.1.0-py3.7.egg/tensorflow_datasets/core/download/download_manager.py in _handle_download_result(self, resource, tmp_dir_path, sha256, dl_size)
215 self._record_sizes_checksums()
216 elif (dl_size, sha256) != self._sizes_checksums.get(resource.url, None):
--> 217 raise NonMatchingChecksumError(resource.url, tmp_path)
218 download_path = self._get_final_dl_path(resource.url, sha256)
219 resource_lib.write_info_file(resource, download_path, self._dataset_name,

NonMatchingChecksumError: Artifact http://pix3d.csail.mit.edu/data/pix3d.zip, downloaded to ./pix3d.csail.mit.edu_pix3d4gh-bf6GMM2oHrBcmQyos_K1PupfuBITgTic9qdZ2Xc.zip.tmp.3b8459f449b34fc68ed246e8c371495f/pix3d.zip, has wrong checksum.`

• Operating System: Ubuntu 18.04
• Python version: 3.7.6
• tensorflow-datasets/tfds-nightly version: 2.1.0

Could you please guide how this error could be resolved ?

[Eshan-Agarwal]

@aradhyamathur are you following instructions provided here properly for implementing dataset ?

Please provide some code

[vijayphoenix]

Use --register_checksum=True while preparing a dataset for the first time

[shikhar2707]

python -m tensorflow_datasets.scripts.download_and_prepare --registrer_checksums=True --datasets='dataset_name'

This command line too will work!

[Eshan-Agarwal]

Yes this is exactly you have to pass when implementing new dataset

[aradhyamathur]

Thanks

[Eshan-Agarwal]

@aradhyamathur If it solves your issue please close this issue