def test_pipeline_first_fit_stage(self):
fit_stages = self.pipeline_event.fitStages
fit_event1 = fit_stages[0].fe
# First Stage
transformer = fit_event1.model
expected_transformer = modeldb_types.Transformer(
-1, 'PCA', 'decomposition PCA')
utils.is_equal_transformer(transformer, expected_transformer, self)
df = fit_event1.df
expected_df = modeldb_types.DataFrame(-1, [
modeldb_types.DataFrameColumn('A', 'int64'),
modeldb_types.DataFrameColumn('B', 'int64'),
], 100, 'digits-dataset')
utils.is_equal_dataframe(df, expected_df, self)
spec = fit_event1.spec
expected_spec = modeldb_types.TransformerSpec(-1, 'PCA', [
modeldb_types.HyperParameter('copy', 'True', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter('n_components', 'None', 'NoneType',
FMIN, FMAX),
modeldb_types.HyperParameter('whiten', 'False', 'bool', FMIN,
FMAX),
], 'decomposition PCA')
utils.is_equal_transformer_spec(spec, expected_spec, self)
self.assertEqual(fit_event1.featureColumns, ['A', 'B'])
def test_pipeline_second_fit_stage(self):
fit_stages = self.pipeline_event.fitStages
fit_event2 = fit_stages[1].fe
# Second Stage
transformer = fit_event2.model
expected_transformer = modeldb_types.Transformer(
-1,
'LinearRegression',
'basic linear reg')
utils.is_equal_transformer(transformer, expected_transformer, self)
df = fit_event2.df
expected_df = modeldb_types.DataFrame(
-1,
[],
100,
'')
utils.is_equal_dataframe(df, expected_df, self)
spec = fit_event2.spec
expected_spec = modeldb_types.TransformerSpec(
-1,
'LinearRegression',
[
modeldb_types.HyperParameter(
'copy_X', 'True', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter(
'normalize', 'False', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter('n_jobs', '1', 'int', FMIN, FMAX),
modeldb_types.HyperParameter(
'fit_intercept', 'True', 'bool', FMIN, FMAX)
],
'basic linear reg')
utils.is_equal_transformer_spec(spec, expected_spec, self)
def test_transformer_spec_construction(self):
spec = self.fit_event.spec
utils.validate_transformer_spec_struct(self.fit_event.spec, self)
expected_spec = modeldb_types.TransformerSpec(
-1,
'LabelEncoder',
[],
'label encoder') # Fix hyperparams.
utils.is_equal_transformer_spec(expected_spec, spec, self)
def test_transformer_spec(self):
spec = self.fit_event.spec
expected_spec = modeldb_types.TransformerSpec(-1, 'LinearRegression', [
modeldb_types.HyperParameter('copy_X', 'True', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter('normalize', 'False', 'bool', FMIN,
FMAX),
modeldb_types.HyperParameter('n_jobs', '1', 'int', FMIN, FMAX),
modeldb_types.HyperParameter('fit_intercept', 'True', 'bool', FMIN,
FMAX)
], 'linear reg')
utils.is_equal_transformer_spec(spec, expected_spec, self)
def test_overall_pipeline_fit_event(self):
fit_event = self.pipeline_event.pipelineFit
utils.validate_fit_event_struct(fit_event, self)
transformer = fit_event.model
expected_transformer = modeldb_types.Transformer(
-1,
'Pipeline',
'pipeline with pca + logistic')
utils.is_equal_transformer(transformer, expected_transformer, self)
df = fit_event.df
expected_df = modeldb_types.DataFrame(
-1,
[
modeldb_types.DataFrameColumn('A', 'int64'),
modeldb_types.DataFrameColumn('B', 'int64'),
],
100,
'digits-dataset')
utils.is_equal_dataframe(df, expected_df, self)
spec = fit_event.spec
expected_spec = modeldb_types.TransformerSpec(
-1,
'Pipeline',
[
modeldb_types.HyperParameter(
'logistic__n_jobs', '1', 'int', FMIN, FMAX),
modeldb_types.HyperParameter(
'pca__copy', 'True', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter(
'pca__n_components', 'None', 'NoneType', FMIN, FMAX),
modeldb_types.HyperParameter(
'logistic__fit_intercept', 'True', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter(
'pca__whiten', 'False', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter(
'steps', "[('pca', PCA(copy=True, n_components=None, whiten=False)), ('logistic', LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False))]", 'list', FMIN, FMAX),
modeldb_types.HyperParameter(
'logistic', 'LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)', 'LinearRegression', FMIN, FMAX),
modeldb_types.HyperParameter(
'pca', 'PCA(copy=True, n_components=None, whiten=False)', 'PCA', FMIN, FMAX),
modeldb_types.HyperParameter(
'logistic__normalize', 'False', 'bool', FMIN, FMAX),
modeldb_types.HyperParameter(
'logistic__copy_X', 'True', 'bool', FMIN, FMAX)
],
'pipeline with pca + logistic')
utils.is_equal_transformer_spec(spec, expected_spec, self)
self.assertItemsEqual(fit_event.featureColumns, ['A', 'B'])