Vowal Wabbit

User 324 | 6/19/2014, 9:12:12 AM

Hello everybody

I have to vw datasets (test.vw, train.vw).

Now I am trying to transform this vw code into graph lab code but not sure how to handle it:

vw train.vw -c -k --passes 40 -l 0.85 -f model.vw --lossfunction quantile --quantiletau 0.6 vw test.vw -t -i model.vw -p shop.preds.txt


User 19 | 6/20/2014, 4:52:31 PM

Assuming you have imported your .vw files into an SFrame (where each namespace of features is a single column of the SFrame), then you should be able to create the above model using:

import graphlab m = graphlab.vowpalwabbit.create(trainsf, targetcolumn='yourtargetcolumn', lossfunction='quantile', maxiterations=40, stepsize=.85, commandlineargs='--quantile_tau .6')

Predictions can be made via

pred = m.predict(test_sf)

If you want to save these to a text file, you can use the SFrame utilities for that.

Hope that helps!