def print_output(library, algorithm, stages, columns, params, functions, times,
accuracy_type, accuracies, data):
if params.output_format == 'csv':
print_header(columns, params)
for i in range(len(accuracies)):
print_row(columns,
params,
prep_function=functions[2 * i],
function=functions[2 * i + 1],
time=times[2 * i],
prep_time=times[2 * i + 1],
accuracy=accuracies[i])
elif params.output_format == 'json':
output = []
output.append({
'library': library,
'algorithm': algorithm,
'input_data': {
'data_format': params.data_format,
'data_order': params.data_order,
'data_type': str(params.dtype),
'dataset_name': params.dataset_name,
'rows': data[0].shape[0],
'columns': data[0].shape[1]
}
})
if hasattr(params, 'n_classes'):
output[-1]['input_data'].update({'classes': params.n_classes})
for i in range(len(stages)):
result = {
'stage': stages[i],
}
if 'daal' in stages[i]:
result.update({
'conversion_to_daal4py': times[2 * i],
'prediction_time': times[2 * i + 1]
})
elif 'train' in stages[i]:
result.update({
'matrix_creation_time': times[2 * i],
'training_time': times[2 * i + 1]
})
else:
result.update({
'matrix_creation_time': times[2 * i],
'prediction_time': times[2 * i + 1]
})
if accuracies[i] is not None:
result.update({f'{accuracy_type}': accuracies[i]})
output.append(result)
print(json.dumps(output, indent=4))
def test_predict(X, X_init):
algorithm = kmeans(fptype=getFPType(X),
nClusters=params.n_clusters,
maxIterations=0,
assignFlag=True,
accuracyThreshold=0.0)
return algorithm.compute(X, X_init)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'n_clusters', 'time')
print_header(columns, params)
# Time fit
fit_time, _ = time_mean_min(test_fit,
X,
X_init,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
print_row(columns, params, function='KMeans.fit', time=fit_time)
# Time predict
predict_time, _ = time_mean_min(test_predict,
X,
X_init,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
print_row(columns, params, function='KMeans.predict', time=predict_time)
except ImportError:
from sklearn.ensemble import RandomForestRegressor
# Load data
X = np.load(params.filex.name)
y = np.load(params.filey.name)
# Create our random forest regressor
regr = RandomForestRegressor(n_estimators=params.num_trees,
max_depth=params.max_depth,
max_features=params.max_features,
random_state=params.seed)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'num_trees', 'time')
params.size = size_str(X.shape)
params.dtype = X.dtype
print_header(columns, params)
# Time fit and predict
fit_time, _ = time_mean_min(regr.fit, X, y,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
print_row(columns, params, function='df_regr.fit', time=fit_time)
predict_time, y_pred = time_mean_min(regr.predict, X,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
print_row(columns, params, function='df_regr.predict', time=predict_time)
# Time fit and predict
fit_time, res = time_mean_min(test_fit,
X,
y,
penalty='l2',
C=params.C,
verbose=params.verbose,
fit_intercept=params.fit_intercept,
tol=params.tol,
max_iter=params.maxiter,
solver=params.solver,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
beta, intercept, solver_result, params.multiclass = res
print_row(columns, params, function='LogReg.fit', time=fit_time)
predict_time, yp = time_mean_min(test_predict,
X,
beta,
intercept=intercept,
multi_class=params.multiclass,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
y_pred = np.argmax(yp, axis=1)
acc = 100 * accuracy_score(y_pred, y)
print_row(columns,
params,
function='LogReg.predict',
time=predict_time,
accuracy=acc)
X.shape[0],
eigenvalues,
eigenvectors,
whiten=params.whiten)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'svd_solver', 'n_components', 'whiten', 'time')
print_header(columns, params)
# Time fit
fit_time, res = time_mean_min(test_fit,
X,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
print_row(columns, params, function='PCA.fit', time=fit_time)
# Time transform
transform_time, tr = time_mean_min(test_transform,
Xp,
*res[:3],
outer_loops=params.transform_outer_loops,
inner_loops=params.transform_inner_loops)
print_row(columns, params, function='PCA.transform', time=transform_time)
if params.write_results:
np.save('pca_daal4py_X.npy', X)
np.save('pca_daal4py_Xp.npy', Xp)
np.save('pca_daal4py_eigvals.npy', res[1])
np.save('pca_daal4py_eigvecs.npy', res[2])
np.save('pca_daal4py_tr.npy', tr)
def main():
parser = argparse.ArgumentParser(description='daal4py SVC benchmark with '
'linear kernel')
parser.add_argument('-x',
'--filex',
'--fileX',
type=argparse.FileType('r'),
required=True,
help='Input file with features, in NPY format')
parser.add_argument('-y',
'--filey',
'--fileY',
type=argparse.FileType('r'),
required=True,
help='Input file with labels, in NPY format')
parser.add_argument('-C',
dest='C',
type=float,
default=0.01,
help='SVM slack parameter')
parser.add_argument('--kernel',
choices=('linear', ),
default='linear',
help='SVM kernel function')
parser.add_argument('--maxiter',
type=int,
default=2000,
help='Maximum iterations for the iterative solver. '
'-1 means no limit.')
parser.add_argument('--max-cache-size',
type=int,
default=64,
help='Maximum cache size, in gigabytes, for SVM.')
parser.add_argument('--tau',
type=float,
default=1e-12,
help='Tau parameter for working set selection scheme')
parser.add_argument('--tol', type=float, default=1e-16, help='Tolerance')
parser.add_argument('--no-shrinking',
action='store_false',
default=True,
dest='shrinking',
help="Don't use shrinking heuristic")
params = parse_args(parser,
loop_types=('fit', 'predict'),
prefix='daal4py')
# Load data and cast to float64
X_train = np.load(params.filex.name).astype('f8')
y_train = np.load(params.filey.name).astype('f8')
cache_size_bytes = get_optimal_cache_size(X_train.shape[0],
max_cache=params.max_cache_size)
params.cache_size_mb = cache_size_bytes / 2**20
params.cache_size_bytes = cache_size_bytes
params.n_classes = np.unique(y_train).size
# This is necessary for daal
y_train[y_train == 0] = -1
y_train = y_train[:, np.newaxis]
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype',
'size', 'kernel', 'cache_size_mb', 'C', 'sv_len', 'n_classes',
'accuracy', 'time')
params.size = size_str(X_train.shape)
params.dtype = X_train.dtype
print_header(columns, params)
# Time fit and predict
fit_time, res = time_mean_min(test_fit,
X_train,
y_train,
params,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
res, support, indices, n_support = res
params.sv_len = support.shape[0]
print_row(columns, params, function='SVM.fit', time=fit_time)
predict_time, yp = time_mean_min(test_predict,
X_train,
res,
params,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
print_row(columns,
params,
function='SVM.predict',
time=predict_time,
accuracy=f'{100*accuracy_score(yp, y_train):.3}')
parser.add_argument('--metrics',
nargs='*',
default=['cosine', 'correlation'],
choices=('cosine', 'correlation'),
help='Metrics to test for pairwise_distances')
params = parse_args(parser,
size=(1000, 150000),
dtypes=('f8', 'f4'),
prefix='daal4py')
# Generate random data
X = np.random.rand(*params.shape).astype(params.dtype)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'time')
print_header(columns, params)
for metric in params.metrics:
pairwise_distances = getattr(daal4py, f'{metric}_distance')
def test_distances(pairwise_distances, X):
algorithm = pairwise_distances(fptype=getFPType(X))
return algorithm.compute(X)
time, _ = time_mean_min(test_distances,
pairwise_distances,
X,
outer_loops=params.outer_loops,
inner_loops=params.inner_loops)
print_row(columns, params, function=metric.capitalize(), time=time)
parser.add_argument('--solver', default='auto',
help='Solver used for training')
params = parse_args(parser, size=(1000000, 50), dtypes=('f8', 'f4'),
loop_types=('fit', 'predict'))
# Generate random data
X = np.random.rand(*params.shape).astype(params.dtype)
Xp = np.random.rand(*params.shape).astype(params.dtype)
y = np.random.rand(*params.shape).astype(params.dtype)
# Create our regression object
regr = Ridge(fit_intercept=params.fit_intercept,
solver=params.solver)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'time')
print_header(columns, params)
# Time fit
fit_time, _ = time_mean_min(regr.fit, X, y,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
print_row(columns, params, function='Ridge.fit', time=fit_time)
# Time predict
predict_time, yp = time_mean_min(regr.predict, Xp,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
print_row(columns, params, function='Ridge.predict', time=predict_time)
max_depth=params.max_depth,
max_features=params.max_features,
random_state=params.seed)
columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
'num_trees', 'n_classes', 'accuracy', 'time')
params.n_classes = len(np.unique(y))
params.size = size_str(X.shape)
params.dtype = X.dtype
print_header(columns, params)
# Time fit and predict
fit_time, _ = time_mean_min(clf.fit,
X,
y,
outer_loops=params.fit_outer_loops,
inner_loops=params.fit_inner_loops)
print_row(columns, params, function='df_clsf.fit', time=fit_time)
predict_time, y_pred = time_mean_min(clf.predict,
X,
outer_loops=params.predict_outer_loops,
inner_loops=params.predict_inner_loops)
acc = 100 * accuracy_score(y_pred, y)
print_row(columns,
params,
function='df_clsf.predict',
time=predict_time,
accuracy=acc)