def test_text_classification_input_string():
assert TextClassificationRecord(
inputs="A text") == TextClassificationRecord(inputs=dict(
text="A text"))
assert TextClassificationRecord(
inputs=["A text", "another text"]) == TextClassificationRecord(
inputs=dict(text=["A text", "another text"]))
def test_metadata_values_length():
text = "oh yeah!"
metadata = {"too_long": "a" * 200}
record = TextClassificationRecord(inputs={"text": text}, metadata=metadata)
assert len(record.metadata["too_long"]) == MAX_KEYWORD_LENGTH
record = TokenClassificationRecord(text=text,
tokens=text.split(),
metadata=metadata)
assert len(record.metadata["too_long"]) == MAX_KEYWORD_LENGTH
def _text_classification_sdk_to_record(
sdk: Union[models.TextClassificationRecord, Dict[str, Any]]
) -> TextClassificationRecord:
"""Transforms and returns the sdk model as a `TextClassificationRecord` record"""
if isinstance(sdk, models.TextClassificationRecord):
sdk = sdk.to_dict()
record = TextClassificationRecord(
id=sdk.get("id"),
event_timestamp=sdk.get("event_timestamp"),
inputs=sdk.get("text", sdk.get("inputs")),
multi_label=sdk.get("multi_label"),
status=sdk.get("status"),
)
prediction = sdk.get("prediction")
if prediction:
record.prediction = [(label["class"], label["confidence"])
for label in prediction["labels"]]
record.prediction_agent = prediction["agent"]
annotation = sdk.get("annotation")
if annotation:
# TODO(dfidalgo): it's depends on multilabel field?
record.annotation = [
label["class"] for label in annotation["labels"]
]
record.annotation_agent = annotation["agent"]
explanation = sdk.get("explanation")
if explanation:
record.explanation = {
key: [
TokenAttributions(**attribution)
for attribution in attributions
]
for key, attributions in explanation
}
metadata = sdk.get("metadata")
if metadata:
record.metadata = metadata
return record
def test_text_classification_record(annotation, status, expected_status):
"""Just testing its dynamic defaults"""
record = TextClassificationRecord(inputs={"text": "test"},
annotation=annotation,
status=status)
assert record.status == expected_status
def test_text_classification_record_none_inputs():
"""Test validation error for None in inputs"""
with pytest.raises(ValidationError):
TextClassificationRecord(inputs={"text": None})