#=============================================================================== # Copyright 2014 Intel Corporation # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. #=============================================================================== import os import sys test_path = os.path.abspath(os.path.dirname(__file__)) unittest_data_path = os.path.join(test_path, "unittest_data") examples_path = os.path.join(os.path.dirname(test_path), "examples", "daal4py") sys.path.insert(0, examples_path) os.chdir(examples_path) import unittest import numpy as np import pandas as pd from scipy.sparse import csr_matrix from daal4py.sklearn._utils import get_daal_version # First item is major version - 2021, # second is minor+patch - 0110, # third item is status - B daal_version = get_daal_version() print('DAAL version:', daal_version) def check_version(rule, target): if not isinstance(rule[0], type(target)): if rule > target: return False else: for rule_item in rule: if rule_item > target: return False if rule_item[0] == target[0]: break return True def check_libraries(rule): for rule_item in rule: try: __import__(rule_item, fromlist=['']) except ImportError: return False return True # function reading file and returning numpy array def np_read_csv(f, c=None, s=0, n=np.iinfo(np.int64).max, t=np.float64): if s == 0 and n == np.iinfo(np.int64).max: return np.loadtxt(f, usecols=c, delimiter=',', ndmin=2, dtype=t) a = np.genfromtxt(f, usecols=c, delimiter=',', skip_header=s, max_rows=n, dtype=t) if a.shape[0] == 0: raise Exception("done") if a.ndim == 1: return a[:, np.newaxis] return a # function reading file and returning pandas DataFrame def pd_read_csv(f, c=None, s=0, n=None, t=np.float64): return pd.read_csv(f, usecols=c, delimiter=',', header=None, skiprows=s, nrows=n, dtype=t) # function reading file and returning scipy.sparse.csr_matrix def csr_read_csv(f, c=None, s=0, n=None, t=np.float64): return csr_matrix(pd_read_csv(f, c, s=s, n=n, t=t)) def add_test(cls, e, f=None, attr=None, ver=(0, 0), req_libs=[]): import importlib @unittest.skipUnless( check_version(ver, daal_version), str(ver) + " not supported in this library version " + str(daal_version) ) @unittest.skipUnless( check_libraries(req_libs), "cannot import required libraries " + str(req_libs) ) def testit(self): ex = importlib.import_module(e) result = self.call(ex) if f and attr: testdata = np_read_csv(os.path.join(unittest_data_path, f)) actual = attr(result) if callable(attr) else getattr(result, attr) self.assertTrue( np.allclose(actual, testdata, atol=1e-05), msg="Discrepancy found: {}".format(np.abs(actual - testdata).max()) ) else: self.assertTrue(True) setattr(cls, 'test_' + e, testit) class Base(): """ We also use generic functions to test these, they get added later. """ def test_svd_batch(self): import svd_batch as ex (data, result) = self.call(ex) self.assertTrue(np.allclose(data, np.matmul( np.matmul(result.leftSingularMatrix, np.diag(result.singularValues[0])), result.rightSingularMatrix) ) ) def test_svd_stream(self): import svd_streaming as ex result = self.call(ex) data = np.loadtxt("./data/distributed/svd_1.csv", delimiter=',') for f in ["./data/distributed/svd_{}.csv".format(i) for i in range(2, 5)]: data = np.append(data, np.loadtxt(f, delimiter=','), axis=0) self.assertTrue(np.allclose(data, np.matmul(np.matmul(result.leftSingularMatrix, np.diag(result.singularValues[0])), result.rightSingularMatrix))) def test_qr_batch(self): import qr_batch as ex (data, result) = self.call(ex) self.assertTrue(np.allclose(data, np.matmul(result.matrixQ, result.matrixR))) def test_qr_stream(self): import qr_streaming as ex result = self.call(ex) data = np.loadtxt("./data/distributed/qr_1.csv", delimiter=',') for f in ["./data/distributed/qr_{}.csv".format(i) for i in range(2, 5)]: data = np.append(data, np.loadtxt(f, delimiter=','), axis=0) self.assertTrue(np.allclose(data, np.matmul(result.matrixQ, result.matrixR))) def test_svm_batch(self): testdata = np_read_csv(os.path.join(unittest_data_path, "svm_batch.csv"), range(1)) import svm_batch as ex (decision_result, _, _) = self.call(ex) left = np.absolute(decision_result - testdata).max() right = np.absolute(decision_result.max() - decision_result.min()) * 0.05 self.assertTrue(left < right) gen_examples = [ ('adaboost_batch', None, None, (2020, 'P', 0)), ('adagrad_mse_batch', 'adagrad_mse_batch.csv', 'minimum'), ('association_rules_batch', 'association_rules_batch.csv', 'confidence'), ('bacon_outlier_batch', 'multivariate_outlier_batch.csv', lambda r: r[1].weights), ('brownboost_batch', None, None, (2020, 'P', 0)), ('correlation_distance_batch', 'correlation_distance_batch.csv', lambda r: [[np.amin(r.correlationDistance)], [np.amax(r.correlationDistance)], [np.mean(r.correlationDistance)], [np.average(r.correlationDistance)]]), ('cosine_distance_batch', 'cosine_distance_batch.csv', lambda r: [[np.amin(r.cosineDistance)], [np.amax(r.cosineDistance)], [np.mean(r.cosineDistance)], [np.average(r.cosineDistance)]]), # ('gradient_boosted_regression_batch', 'gradient_boosted_regression_batch.csv', # lambda x: x[1].prediction), ('cholesky_batch', 'cholesky_batch.csv', 'choleskyFactor'), ('covariance_batch', 'covariance.csv', 'covariance'), ('covariance_streaming', 'covariance.csv', 'covariance'), ('decision_forest_classification_default_dense_batch', None, lambda r: r[1].prediction, (2023, 'P', 1)), ('decision_forest_classification_hist_batch', None, lambda r: r[1].prediction, (2023, 'P', 1)), ('decision_forest_regression_default_dense_batch', 'decision_forest_regression_batch.csv', lambda r: r[1].prediction, (2023, 'P', 1)), ('decision_forest_regression_hist_batch', 'decision_forest_regression_batch.csv', lambda r: r[1].prediction, (2023, 'P', 1)), ('decision_forest_regression_default_dense_batch', 'decision_forest_regression_batch_20230101.csv', lambda r: r[1].prediction, (2023, 'P', 101)), ('decision_forest_regression_hist_batch', 'decision_forest_regression_batch_20230101.csv', lambda r: r[1].prediction, (2023, 'P', 101)), ('decision_tree_classification_batch', 'decision_tree_classification_batch.csv', lambda r: r[1].prediction), ('decision_tree_regression_batch', 'decision_tree_regression_batch.csv', lambda r: r[1].prediction), ('distributions_bernoulli_batch',), ('distributions_normal_batch',), ('distributions_uniform_batch',), ('em_gmm_batch', 'em_gmm.csv', lambda r: r.covariances[0]), ('gbt_cls_model_create_from_lightgbm_batch', None, None, ((2020, 'P', 2), (2021, 'B', 109)), ['lightgbm']), ('gbt_cls_model_create_from_xgboost_batch', None, None, ((2020, 'P', 2), (2021, 'B', 109)), ['xgboost']), ('gbt_cls_model_create_from_catboost_batch', None, None, (2021, 'P', 4), ['catboost']), ('gradient_boosted_classification_batch',), ('gradient_boosted_regression_batch',), ('implicit_als_batch', 'implicit_als_batch.csv', 'prediction'), ('kdtree_knn_classification_batch', None, None), ('kmeans_batch', 'kmeans_batch.csv', 'centroids'), ('lbfgs_cr_entr_loss_batch', 'lbfgs_cr_entr_loss_batch.csv', 'minimum'), ('lbfgs_mse_batch', 'lbfgs_mse_batch.csv', 'minimum'), ('linear_regression_batch', 'linear_regression_batch.csv', lambda r: r[1].prediction), ('linear_regression_streaming', 'linear_regression_batch.csv', lambda r: r[1].prediction), # return when Logistic Regression will be fixed # ('log_reg_binary_dense_batch', 'log_reg_binary_dense_batch.csv', # lambda r: r[1].prediction), ('log_reg_binary_dense_batch', None, None), ('log_reg_dense_batch',), ('logitboost_batch', None, None, (2020, 'P', 0)), ('low_order_moms_dense_batch', 'low_order_moms_dense_batch.csv', lambda r: np.vstack((r.minimum, r.maximum, r.sum, r.sumSquares, r.sumSquaresCentered, r.mean, r.secondOrderRawMoment, r.variance, r.standardDeviation, r.variation))), ('low_order_moms_streaming', 'low_order_moms_dense_batch.csv', lambda r: np.vstack((r.minimum, r.maximum, r.sum, r.sumSquares, r.sumSquaresCentered, r.mean, r.secondOrderRawMoment, r.variance, r.standardDeviation, r.variation))), ('multivariate_outlier_batch', 'multivariate_outlier_batch.csv', lambda r: r[1].weights), ('naive_bayes_batch', 'naive_bayes_batch.csv', lambda r: r[0].prediction), ('naive_bayes_streaming', 'naive_bayes_batch.csv', lambda r: r[0].prediction), ('normalization_minmax_batch', 'normalization_minmax.csv', 'normalizedData'), ('normalization_zscore_batch', 'normalization_zscore.csv', 'normalizedData'), ('pca_batch', 'pca_batch.csv', 'eigenvectors'), ('pca_transform_batch', 'pca_transform_batch.csv', lambda r: r[1].transformedData), ('pivoted_qr_batch', 'pivoted_qr.csv', 'matrixR'), ('quantiles_batch', 'quantiles.csv', 'quantiles'), ('ridge_regression_batch', 'ridge_regression_batch.csv', lambda r: r[0].prediction), ('ridge_regression_streaming', 'ridge_regression_batch.csv', lambda r: r[0].prediction), ('saga_batch', None, None, (2019, 'P', 3)), ('sgd_logistic_loss_batch', 'sgd_logistic_loss_batch.csv', 'minimum'), ('sgd_mse_batch', 'sgd_mse_batch.csv', 'minimum'), ('sorting_batch',), ('stump_classification_batch', None, None, (2020, 'P', 0)), ('stump_regression_batch', None, None, (2020, 'P', 0)), ('svm_multiclass_batch', 'svm_multiclass_batch.csv', lambda r: r[0].prediction), ('univariate_outlier_batch', 'univariate_outlier_batch.csv', lambda r: r[1].weights), ('dbscan_batch', 'dbscan_batch.csv', 'assignments', (2019, 'P', 5)), ('lasso_regression_batch', None, None, (2019, 'P', 5)), ('elastic_net_batch', None, None, ((2020, 'P', 1), (2021, 'B', 105))), ] for example in gen_examples: add_test(Base, *example) class TestExNpyArray(Base, unittest.TestCase): """ We run and validate all the examples but read data with numpy, so working natively on a numpy arrays. """ def call(self, ex): return ex.main(readcsv=np_read_csv) class TestExPandasDF(Base, unittest.TestCase): """ We run and validate all the examples but read data with pandas, so working natively on a pandas DataFrame """ def call(self, ex): return ex.main(readcsv=pd_read_csv) class TestExCSRMatrix(Base, unittest.TestCase): """ We run and validate all the examples but use scipy-sparse-csr_matrix as input data. We also let algos use CSR method (some algos ignore the method argument since they do not specifically support CSR). """ def call(self, ex): # some algos do not support CSR matrices if ex.__name__.startswith('sorting'): self.skipTest("not supporting CSR") if any(ex.__name__.startswith(x) for x in ['adaboost', 'brownboost', 'stump_classification', 'gbt_cls_model_create', 'decision_forest']): self.skipTest("not supporting CSR") method = \ 'singlePassCSR' if any(x in ex.__name__ for x in ['low_order_moms', 'covariance']) \ else 'fastCSR' # cannot use fastCSR ofr implicit als if 'implicit_als' in ex.__name__: method = 'defaultDense' # kmeans have no special method for CSR if 'kmeans' in ex.__name__: method = 'randomDense' if hasattr(ex, 'dflt_method'): method = ex.dflt_method.replace('defaultDense', 'fastCSR').replace('Dense', 'CSR') return ex.main(readcsv=csr_read_csv, method=method) if __name__ == '__main__': unittest.main()