Benchmarking with gil and with nogil

benchmark.py

import pylab as pl
import numpy as np
from sklearn.linear_model import *
from sklearn.datasets import make_regression
import time


def plot():
    pl.figure("Benchmark with nogil")

    n_alphas = [5, 10, 100, 500]
    l1_ratios = [1, 2, 5, 10]
    n_samples = [200, 500]
    n_features = [200, 500]
    temp = 221
    colors = ["b-", "g-", "r-", "c-"]

    for i, sample in enumerate(n_samples):
    	for j, feature in enumerate(n_features):
    	    X, y = make_regression(n_samples=sample, n_features=feature)
    	    pl.subplot(temp)
    	    temp += 1
    	    for alpha, color in zip(n_alphas, colors):
    	        times1= []
    		for l1_ratio in l1_ratios:
    		    n_l1_ratios = np.linspace(0.1, 1, num=l1_ratio)
    		    clf = ElasticNetCV()
                    t = time.time()
                    clf.fit(X, y)
                    times1.append(time.time() - t)
             pl.plot(l1_ratios, times1, color, label=str(alpha))
    	     pl.legend(loc='upper right')
    	     pl.title("n_samples=%d, n_features=%d" % (sample, feature))

    pl.xlabel('number of l1_ratios')
    pl.ylabel('Time (s)')
    pl.axis('tight')
    #pl.show()
    pl.savefig("withnogil.png")


if __name__ == '__main__':
    plot()