def default_output_file_loaders(cls) -> dict[FileType, FileLoader]:
target_field_names = ["predict", "predictions", "true_rank"]
return {
FileType.json:
JSONFileLoader([
FileLoaderField("predict", target_field_names[0], str),
FileLoaderField("predictions", target_field_names[1], list),
FileLoaderField("true_rank", target_field_names[2], int),
])
}
def default_output_file_loaders(cls) -> dict[FileType, FileLoader]:
return {
FileType.conll:
CoNLLFileLoader([FileLoaderField(1, "pred_tags", str)]),
FileType.json:
JSONFileLoader([
FileLoaderField("tokens", "tokens", list),
FileLoaderField("predicted_tags", "pred_tags", list),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
return {
FileType.json:
JSONFileLoader([
FileLoaderField('text', 'text', str),
FileLoaderField('edits', 'edits', dict),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField('text', 'text', str),
FileLoaderField('edits', 'edits', dict),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
field_names = ["tokens", "true_tags"]
return {
FileType.conll:
CoNLLFileLoader([
FileLoaderField(0, field_names[0], str),
FileLoaderField(1, field_names[1], str),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField("tokens", field_names[0], list),
FileLoaderField("tags", field_names[1], list),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
field_name = 'text'
return {
FileType.text:
TextFileLoader(),
FileType.json:
JSONFileLoader([
FileLoaderField("text", field_name, str),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField("text", field_name, str),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
file_path = cache_api.cache_online_file(
'http://phontron.com/download/explainaboard/pre_computed/kg/entity2wikidata.json', # noqa
'pre_computed/kg/entity2wikidata.json',
)
with open(file_path, 'r') as file:
entity_dic = json.loads(file.read())
map_preprocessor = KGMapPreprocessor(
resources={"dictionary": entity_dic})
target_field_names = [
"true_head",
"true_head_decipher",
"true_link",
"true_tail",
"true_tail_decipher",
]
return {
FileType.json:
JSONFileLoader([
FileLoaderField("gold_head", target_field_names[0], str),
FileLoaderField("gold_head",
target_field_names[1],
str,
parser=map_preprocessor),
FileLoaderField("gold_predicate", target_field_names[2], str),
FileLoaderField("gold_tail", target_field_names[3], str),
FileLoaderField("gold_tail",
target_field_names[4],
str,
parser=map_preprocessor),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField("head", target_field_names[0], str),
FileLoaderField("head",
target_field_names[1],
str,
parser=map_preprocessor),
FileLoaderField("link", target_field_names[2], str),
FileLoaderField("tail", target_field_names[3], str),
FileLoaderField("tail",
target_field_names[4],
str,
parser=map_preprocessor),
]),
}
def default_output_file_loaders(cls) -> dict[FileType, FileLoader]:
return {
FileType.json:
JSONFileLoader([
FileLoaderField("predicted_answers", "predicted_answers", dict)
])
}
def test_tsv_validation(self):
self.assertRaises(
ValueError,
lambda: TSVFileLoader(
[FileLoaderField("0", "field0", str)], use_idx_as_id=True
),
)
def default_output_file_loaders(cls) -> dict[FileType, FileLoader]:
field_name = "hypothesis"
return {
FileType.text:
TextFileLoader(field_name, str),
FileType.json:
JSONFileLoader([FileLoaderField(field_name, field_name, str)]),
}
def test_conll_loader(self):
tabs_path = os.path.join(test_artifacts_path, "ner", "dataset.tsv")
spaces_path = os.path.join(test_artifacts_path, "ner",
"dataset-space.tsv")
loader_true = CoNLLFileLoader([
FileLoaderField(0, 'tokens', str),
FileLoaderField(1, 'true_tags', str),
])
loader_pred = CoNLLFileLoader([
FileLoaderField(1, 'pred_tags', str),
])
tabs_true = loader_true.load(tabs_path, Source.local_filesystem)
spaces_true = loader_true.load(spaces_path, Source.local_filesystem)
self.assertEqual(tabs_true, spaces_true)
tabs_pred = loader_pred.load(tabs_path, Source.local_filesystem)
spaces_pred = loader_pred.load(spaces_path, Source.local_filesystem)
self.assertEqual(tabs_pred, spaces_pred)
def _statistics_func(self, samples: Iterator, sys_info: SysOutputInfo):
if sys_info.source_language is None or sys_info.target_language is None:
raise ValueError(
'source or target languages must be specified to load '
f'translation data, but source={sys_info.source_language} '
f', target={sys_info.target_language}'
)
src = FileLoaderField(('translation', sys_info.source_language), '', str)
trg = FileLoaderField(('translation', sys_info.target_language), '', str)
return {
'source_vocab': accumulate_vocab_from_samples(
samples,
lambda x: FileLoader.find_field(x, src),
unwrap(sys_info.source_tokenizer),
),
'target_vocab': accumulate_vocab_from_samples(
samples,
lambda x: FileLoader.find_field(x, trg),
unwrap(sys_info.target_tokenizer),
),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
target_field_names = ["context", "options", "question_mark", "answers"]
return {
FileType.json:
JSONFileLoader([
FileLoaderField("context", target_field_names[0], str),
FileLoaderField("options", target_field_names[1], list),
FileLoaderField("question_mark", target_field_names[2], str),
FileLoaderField("answers", target_field_names[3], dict),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField("context", target_field_names[0], str),
FileLoaderField("options", target_field_names[1], list),
FileLoaderField("question_mark", target_field_names[2], str),
FileLoaderField("answers", target_field_names[3], dict),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
target_names = ["text1", "text2", "true_label"]
return {
FileType.tsv:
TSVFileLoader([
FileLoaderField(0, target_names[0], str),
FileLoaderField(1, target_names[1], str),
FileLoaderField(2, target_names[2], str),
], ),
FileType.json:
JSONFileLoader([
FileLoaderField("text1", target_names[0], str),
FileLoaderField("text2", target_names[1], str),
FileLoaderField("true_label", target_names[2], str),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField("text1", target_names[0], str),
FileLoaderField("text2", target_names[1], str),
FileLoaderField("label", target_names[2], str),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
field_names = ["aspect", "text", "true_label"]
return {
FileType.tsv:
TSVFileLoader([
FileLoaderField(0, field_names[0], str),
FileLoaderField(1, field_names[1], str),
FileLoaderField(2, field_names[2], str),
], ),
FileType.json:
JSONFileLoader([
FileLoaderField(field_names[0], field_names[0], str),
FileLoaderField(field_names[1], field_names[1], str),
FileLoaderField(field_names[2], field_names[2], str),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField(field_names[0], field_names[0], str),
FileLoaderField(field_names[1], field_names[1], str),
FileLoaderField(field_names[2], field_names[2], str),
]),
}
def test_load_in_memory_tsv(self):
loader = Loader(
dataset_data=load_file_as_str(self.dataset),
output_data=load_file_as_str(
os.path.join(test_artifacts_path, "text_classification",
"output.txt")),
dataset_source=Source.in_memory,
output_source=Source.in_memory,
dataset_file_type=FileType.tsv,
output_file_type=FileType.text,
dataset_file_loader=TSVFileLoader(
[FileLoaderField(0, "field0", str)], use_idx_as_id=True),
output_file_loader=TextFileLoader("output", str),
)
samples = [sample for sample in loader.load()]
self.assertEqual(len(samples), 10)
self.assertEqual(set(samples[0].keys()), {"id", "field0", "output"})
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
return {
FileType.tsv:
TSVFileLoader([
FileLoaderField(0, cls.OUTPUT_FIELDS[0], str),
FileLoaderField(1, cls.OUTPUT_FIELDS[1], str),
], ),
FileType.json:
JSONFileLoader([
FileLoaderField(cls.JSON_FIELDS[0], cls.OUTPUT_FIELDS[0], str),
FileLoaderField(cls.JSON_FIELDS[1], cls.OUTPUT_FIELDS[1], str),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField(cls.JSON_FIELDS[0], cls.OUTPUT_FIELDS[0], str),
FileLoaderField(cls.JSON_FIELDS[1], cls.OUTPUT_FIELDS[1], str),
]),
}
def default_dataset_file_loaders(cls) -> dict[FileType, FileLoader]:
target_field_names = ["context", "question", "answers"]
return {
FileType.json:
JSONFileLoader([
FileLoaderField(
target_field_names[0],
target_field_names[0],
str,
strip_before_parsing=False,
),
FileLoaderField(
target_field_names[1],
target_field_names[1],
str,
strip_before_parsing=False,
),
FileLoaderField(target_field_names[2], target_field_names[2]),
]),
FileType.datalab:
DatalabFileLoader([
FileLoaderField(
"context",
target_field_names[0],
str,
strip_before_parsing=False,
),
FileLoaderField(
"question",
target_field_names[1],
str,
strip_before_parsing=False,
),
FileLoaderField("answers", target_field_names[2]),
]),
}
def default_output_file_loaders(cls) -> dict[FileType, FileLoader]:
field_name = "predicted_edits"
return {
FileType.json:
JSONFileLoader([FileLoaderField(field_name, field_name, dict)]),
}
def test_text_file_loader_validate(self):
loader = TextFileLoader(target_name="prediction", dtype=int)
self.assertRaises(
ValueError,
lambda: loader.add_fields([FileLoaderField("test", "test", str)]),
)
