Module: tff.simulation.datasets.gldv2

Libraries for the federated Google Landmark v2 dataset for simulation.

Classes

class ClientData: Object to hold a federated dataset.

class FilePerUserClientData: A tf.simulation.ClientData that maps a set of files to a dataset.

Functions

load_data(...): Loads a federated version of the Google Landmark v2 dataset.

FED_GLD_CACHE 'gld160k'
FED_GLD_SPLIT_FILE_BUNDLE 'landmarks-user-160k'
FED_GLD_SPLIT_FILE_BUNDLE_MD5_CHECKSUM '53c36bd7d5fc12f927af2820b7e4a57c'
FED_GLD_SPLIT_FILE_DOWNLOAD_URL 'http://storage.googleapis.com/gresearch/federated-vision-datasets/landmarks-user-160k.zip'
FED_GLD_TEST_SPLIT_FILE 'test.csv'
FED_GLD_TRAIN_SPLIT_FILE 'federated_train.csv'
GLD_SHARD_BASE_URL 'https://s3.amazonaws.com/google-landmark'
KEY_CLASS 'class'
KEY_IMAGE_BYTES 'image/encoded_jpeg'
KEY_IMAGE_DECODED 'image/decoded'
LOGGER 'gldv2'
MINI_GLD_CACHE 'gld23k'
MINI_GLD_TEST_DOWNLOAD_URL 'https://storage.googleapis.com/tff-datasets-public/mini_gld_test.csv'
MINI_GLD_TEST_SPLIT_FILE 'mini_gld_test.csv'
MINI_GLD_TEST_SPLIT_FILE_MD5_CHECKSUM '298e9d19d66357236f66fe8e22920933'
MINI_GLD_TRAIN_DOWNLOAD_URL 'https://storage.googleapis.com/tff-datasets-public/mini_gld_train_split.csv'
MINI_GLD_TRAIN_SPLIT_FILE 'mini_gld_train_split.csv'
MINI_GLD_TRAIN_SPLIT_FILE_MD5_CHECKSUM '9fd62cf79a67046fdd673d3a0ac52841'
NUM_SHARD_TRAIN 500
TEST_FILE_NAME 'test.tfRecord'
TRAIN_SUB_DIR 'train'