GLDV2 - Google Landmarks Download Issue

[sarapieri]

GLDV2 Dataset

I am trying to use the load_data function on the GLDV2 dataset in the reduced version of 23K (see function call below) using this code:

import tensorflow as tf
import tensorflow_federated as tff

gldv2_train, gldv2_test = tff.simulation.datasets.gldv2.load_data(gld23k = True)

Environment:

• Linux
• TensorFlow Federated: 0.56.0
• TensorFlow: 2.12.0
• Python version: 3.9
• Not building it from source

Expected behavior
Download the data from landmarks-user-160k.zip

Error
raise ValueError(
ValueError: Incomplete or corrupted file detected. The md5 file hash does not match the provided value of 825975950b2e22f0f66aa8fd26c1f153 images_000.tar

[edvinasstaupas]

Having seemingly the same problem
Using Notebook 2023/09/22 version

gldv2_train, gldv2_test = tff.simulation.datasets.gldv2.load_data() produces:

Incomplete or corrupted file detected. The md5 file hash does not match the provided value of 825975950b2e22f0f66aa8fd26c1f153  images_000.tar

and

gldv2_train, gldv2_test = tff.simulation.datasets.gldv2.load_data(num_worker=5, cache_dir='/cache') produces:

[/usr/local/lib/python3.10/dist-packages/tensorflow_federated/python/simulation/datasets/gldv2.py](https://localhost:8080/#) in load_data(num_worker, cache_dir, gld23k, base_url)
    450     logger.info('Try loading dataset from cache')
--> 451     return vision_datasets_utils.load_data_from_cache(
    452         existing_data_cache, TRAIN_SUB_DIR, TEST_FILE_NAME, LOGGER

11 frames
NotFoundError: Could not find directory /cache/gld160k/train

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
[/usr/local/lib/python3.10/dist-packages/keras/src/utils/data_utils.py](https://localhost:8080/#) in get_file(fname, origin, untar, md5_hash, file_hash, cache_subdir, hash_algorithm, extract, archive_format, cache_dir)
    360         if os.path.exists(fpath) and file_hash is not None:
    361             if not validate_file(fpath, file_hash, algorithm=hash_algorithm):
--> 362                 raise ValueError(
    363                     "Incomplete or corrupted file detected. "
    364                     f"The {hash_algorithm} "

ValueError: Incomplete or corrupted file detected. The md5 file hash does not match the provided value of f62fb3b43041a320d474b6dfbc6696af  images_025.tar
.

@sarapieri have you found anyway to fix this?