I have a simple Python code (digit recognition exercise from Kaggle), which runs fine if I execute it from the command line (I use Windows 8.1 64-bit with Enthought Canopy 1.4.1).
import numpy
from sklearn.ensemble import RandomForestClassifier
from sklearn import cross_validation
print "\nreading training data..."
dataFilename = "D:\\Kaggle\\Digit Recognizer\\Data\\train.csv"
dataFile = open(dataFilename, 'r')
data = numpy.array([map(int, line.replace('\n', '').split(',')) for line in dataFile.readlines()[1:]])
dataFile.close()
print "\nseparating training data to features and targets..."
# use all data to train the algorithm
trainingSet_Y = data[:, 0]
trainingSet_X = data[:, 1:]
print "\ntraining a classifier..."
classify_RF = RandomForestClassifier(n_estimators = 100, n_jobs = -1)
classify_RF.fit(trainingSet_X, trainingSet_Y)
print "\ncalculating cross-validation score..."
scores = cross_validation.cross_val_score(classify_RF, trainingSet_X, trainingSet_Y, cv = 5, n_jobs = -1)
print("Accuracy: %0.2f (+/- %0.2f)" % (scores.mean(), numpy.std(scores) * 2))
I decided to migrate all my developments to Visual Studio, so I installed Python Tools for Visual Studio 2.1 to start coding/running directly from within VS Community 2013. (Note: order of installation: (1) Canopy, (2) VS2013, and (3) PTVS.)
However, the same code behaves extremely strange when executed from within VS2013. It runs until the cross-validation step, and then it starts to loop over the code and rerun everything over and over, sometime spitting out error messages on its way, as shown below:
As you can see, once it reaches the cross-validation step, it starts over from the beginning, and goes over the code randomly to execute only certain parts of it!
Any ideas?
