def main():
argParser = argparse.ArgumentParser(
prog="svm-linear.py",
description="sklearn two-class SVC benchmark for linear kernel",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
args = getArguments(argParser)
numThreads, daal_version = bench.prepare_benchmark(args)
# This is so the .csv file can go directly into excel. However, it requries adding timing
# in sklearn/daal4sklearn/svm.py around pydaal compute and post-processing
if args.verbose:
print('@ {fit_samples: ' + str(args.fit_samples) +
', fit_repetitions: ' + str(args.fit_repetitions) +
', predict_samples: ' + str(args.predict_samples) +
', predict_repetitions: ' + str(args.predict_repetitions) +
', pyDAAL: ' + str(daal_version) + '}',
file=sys.stderr)
# Load data and cast to float64
X_train = np.load(args.fileX.name).astype('f8')
y_train = np.load(args.fileY.name).astype('f8')
v, f = X_train.shape
cache_size_bytes = getOptimalCacheSize(X_train.shape[0])
cache_size_mb = cache_size_bytes / 1024**2
meta_info = ",".join([
args.prefix, 'SVM',
str(numThreads),
str(v),
str(f),
str(int(cache_size_mb))
])
svc_params_dict = {
'C': 0.01,
'kernel': 'linear',
'max_iter': 2000,
'cache_size': cache_size_mb,
'tol': 1e-16,
'shrinking': True
}
if args.verbose:
print("@ {}".format(svc_params_dict), file=sys.stderr)
if args.header:
print(
'prefix_ID,function,threads,rows,features,cache-size-MB,fit,predict,accuracy,sv-len,classes'
)
_bench(meta_info, X_train, y_train, args.fit_samples, args.fit_repetitions,
args.predict_samples, args.predict_repetitions, svc_params_dict)
type=int,
dest="num_threads",
default=0,
help="Number of threads for DAAL to use")
args = argParser.parse_args()
return args
argParser = argparse.ArgumentParser(
prog="df_clsf_bench.py",
description="Execute RandomForest classification",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
args = getArguments(argParser)
num_threads, daal_version = bench.prepare_benchmark(args)
import sklearn
try:
from daal4py.sklearn.ensemble import RandomForestClassifier as rfClassifier
except ImportError:
from sklearn.ensemble import RandomForestClassifier as rfClassifier
if args.use_sklearn_class:
from sklearn.ensemble import RandomForestClassifier as rfClassifier
import timeit
if args.fileX is None or args.fileY is None:
argParser.error(
"Please specify data for the algorithm to train on. Use --fileX and --fileY or --generate options."
from __future__ import print_function
import numpy as np
import timeit
from numpy.random import rand
from sklearn import linear_model
from args import getArguments, coreString
import sklearn
import bench
import argparse
argParser = argparse.ArgumentParser(prog="ridge.py",
description="sklearn ridge regression benchmark",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
args = getArguments(argParser)
REP = args.iteration if args.iteration != '?' else 10
core_number, daal_version = bench.prepare_benchmark(args)
def st_time(func):
def st_func(*args, **keyArgs):
times = []
for n in range(REP):
t1 = timeit.default_timer()
r = func(*args, **keyArgs)
t2 = timeit.default_timer()
times.append(t2-t1)
print (min(times))
return r
return st_func
def main():
argParser = argparse.ArgumentParser(
prog="svm-linear.py",
description="SVC benchmark for linear kernel",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
args = getArguments(argParser)
num_threads, daal_version = prepare_benchmark(args)
# This is so the .csv file can go directly into excel. However, it requries adding timing
# in sklearn/daal4sklearn/svm.py around pydaal compute and post-processing
if args.verbose:
print('@ {fit_samples: ' + str(args.fit_samples) +
', fit_repetitions: ' + str(args.fit_repetitions) +
', predict_samples: ' + str(args.predict_samples) +
', predict_repetitions: ' + str(args.predict_repetitions) +
', pyDAAL: ' + str(daal_version) + '}',
file=sys.stderr)
# Load data and cast to float64
X_train = np.load(args.fileX.name).astype('f8')
y_train = np.load(args.fileY.name).astype('f8')
n_classes = np.unique(y_train).size
y_train[y_train == 0] = -1
y_train = y_train[:, np.newaxis]
v, f = X_train.shape
cache_size = getOptimalCacheSize(X_train.shape[0])
meta_info = ",".join([
args.prefix, 'SVM',
str(num_threads),
str(v),
str(f),
str(int(cache_size))
])
svc_params_dict = {
'C': 0.01,
'maxIterations': 2000,
'tau': 1e-12,
'cacheSize': cache_size,
'accuracyThreshold': 1e-16,
'doShrinking': True,
}
if args.verbose:
print("@ {}".format(svc_params_dict), file=sys.stderr)
if args.header:
print(
'prefix_ID,function,threads,rows,features,cache-size-MB,fit,predict,accuracy,sv_len,classes'
)
bench(meta_info,
X_train,
y_train,
args.fit_samples,
args.fit_repetitions,
args.predict_samples,
args.predict_repetitions,
n_classes,
cache_size,
accuracy_threshold=1e-16,
max_iterations=2000)
