def engineer_features(input_dir, output_dir):
"""Engineer features"""
X_df, y_df = api.load_data(input_dir=input_dir)
result = api.engineer_features()
pipeline, encoder = result.pipeline, result.encoder
X_ft = pipeline.transform(X_df)
y_ft = encoder.transform(y_df)
save_features(X_ft, output_dir)
save_targets(y_ft, output_dir)
def main(input_dir, output_dir):
"""Engineer features"""
import ballet.util.log
ballet.util.log.enable(logger=logger, level='INFO', echo=False)
ballet.util.log.enable(logger=ballet.util.log.logger, level='INFO',
echo=False)
X_df, y_df = load_data(input_dir=input_dir)
out = build()
mapper_X = out.mapper_X
encoder_y = out.encoder_y
X_ft = mapper_X.transform(X_df)
y_ft = encoder_y.transform(y_df)
save_features(X_ft, output_dir)
save_targets(y_ft, output_dir)
def engineer_features(input_dir, output_dir, train_dir):
"""Engineer features"""
# load pipeline trained on development set
if train_dir is not None:
X_df_tr, y_df_tr = api.load_data(input_dir=train_dir)
result = api.engineer_features(X_df_tr, y_df_tr)
else:
result = api.engineer_features()
pipeline, encoder = result.pipeline, result.encoder
# load input data
X_df, y_df = api.load_data(input_dir=input_dir)
# transform entities and targets
X_ft = pipeline.transform(X_df)
y_ft = encoder.transform(y_df)
# save to output dir
save_features(X_ft, output_dir)
save_targets(y_ft, output_dir)