def test_not_explainable(test_dir, data_frame):
label_col = 'label'
df = data_frame(n_samples=100, label_col=label_col)
data_encoder_cols = [BowEncoder('features')]
label_encoder_cols = [CategoricalEncoder(label_col)]
data_cols = [BowFeaturizer('features')]
output_path = os.path.join(test_dir, "tmp", "out")
imputer = Imputer(data_featurizers=data_cols,
label_encoders=label_encoder_cols,
data_encoders=data_encoder_cols,
output_path=output_path).fit(train_df=df, num_epochs=1)
assert not imputer.is_explainable
try:
imputer.explain('some label')
raise pytest.fail(
'imputer.explain should fail with an appropriate error message')
except ValueError as exception:
assert exception.args[0] == 'No explainable data encoders available.'
instance = pd.Series({'features': 'some feature text'})
try:
imputer.explain_instance(instance)
raise pytest.fail(
'imputer.explain_instance should fail with an appropriate error message'
)
except ValueError as exception:
assert exception.args[0] == 'No explainable data encoders available.'
def test_explain_instance_without_label(test_dir, data_frame):
label_col = 'label'
df = data_frame(n_samples=100, label_col=label_col)
data_encoder_cols = [TfIdfEncoder('features')]
label_encoder_cols = [CategoricalEncoder(label_col)]
data_cols = [BowFeaturizer('features')]
output_path = os.path.join(test_dir, "tmp", "out")
imputer = Imputer(data_featurizers=data_cols,
label_encoders=label_encoder_cols,
data_encoders=data_encoder_cols,
output_path=output_path).fit(train_df=df, num_epochs=1)
assert imputer.is_explainable
instance = pd.Series({'features': 'some feature text'})
# explain_instance should not raise an exception
_ = imputer.explain_instance(instance)
assert True
