DEFAULT_OUTPUT_FEATURES = {
"inputs": Feature(vocabulary=get_default_vocabulary(), add_eos=True),
"targets": Feature(vocabulary=get_default_vocabulary(), add_eos=True)
}
# ==================================== C4 ======================================
_c4_config_suffixes = ["", ".noclean", ".realnewslike", ".webtextlike"]
for config_suffix in _c4_config_suffixes:
TaskRegistry.add(
"c4{name}_v020_unsupervised".format(
name=config_suffix.replace(".", "_")),
TfdsTask,
tfds_name="c4/en{config}:2.2.0".format(config=config_suffix),
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
"inputs": None,
"targets": "text"
}),
token_preprocessor=preprocessors.unsupervised,
output_features=DEFAULT_OUTPUT_FEATURES,
metric_fns=[])
# ================================ Wikipedia ===================================
TaskRegistry.add("wikipedia_20190301.en_v003_unsupervised",
TfdsTask,
tfds_name="wikipedia/20190301.en:1.0.0",
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
"inputs": None,
"targets": "text"
DEFAULT_OUTPUT_FEATURES = {
"inputs": Feature(vocabulary=get_default_vocabulary(), add_eos=True),
"targets": Feature(vocabulary=get_default_vocabulary(), add_eos=True)
}
# ==================================== C4 ======================================
# Configurable tasks used for comparisons in Raffel et al., 2019.
_c4_config_suffixes = ["", ".noclean", ".realnewslike", ".webtextlike"]
for config_suffix in _c4_config_suffixes:
TaskRegistry.add(
"c4{name}_v020_unsupervised".format(
name=config_suffix.replace(".", "_")),
TfdsTask,
tfds_name="c4/en{config}:2.2.0".format(config=config_suffix),
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
"inputs": None,
"targets": "text"
}),
token_preprocessor=preprocessors.unsupervised,
output_features=DEFAULT_OUTPUT_FEATURES,
metric_fns=[])
# Final pretraining task used in Raffel et al., 2019.
TaskRegistry.add(
"c4_v220_span_corruption",
TfdsTask,
tfds_name="c4/en:2.2.0".format(config=config_suffix),
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
"inputs": None,
from t5.data import preprocessors
from t5.data.utils import DEFAULT_SPM_PATH
from t5.data.utils import TaskRegistry
from t5.data.utils import TfdsTask
from t5.evaluation import metrics
# ==================================== C4 ======================================
_c4_config_suffixes = ["", ".noclean", ".realnewslike", ".webtextlike"]
for config_suffix in _c4_config_suffixes:
TaskRegistry.add(
"c4{name}_v020_unsupervised".format(
name=config_suffix.replace(".", "_")),
TfdsTask,
tfds_name="c4/en{config}:1.0.0".format(config=config_suffix),
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
"inputs": None,
"targets": "text"
}),
token_preprocessor=preprocessors.unsupervised,
sentencepiece_model_path=DEFAULT_SPM_PATH,
metric_fns=[])
# ================================ Wikipedia ===================================
TaskRegistry.add(
"wikipedia_20190301.en_v003_unsupervised",
TfdsTask,
# 0.0.4 is identical to 0.0.3 except empty records removed.
tfds_name="wikipedia/20190301.en:0.0.4",
text_preprocessor=functools.partial(preprocessors.rekey,
key_map={
