luigi workflows to evaluate models trained by vowpal wabbit.

If you'd like to use vw-luigi, you need to install vowpal wabbit and some python modules.

See https://github.com/JohnLangford/vowpal_wabbit.
If you use OSX, you can install vowpal-wabbit through homebrew.

brew install vowpal-wabbit

Workflows in vw-luigi depend on luigi, numpy and scikit-learn. You can install required modules through pip.

pip install -r requirements.txt

In case you use /tmp/work/space/train.vw as training data, /tmp/work/space/test.vw as test data and squared loss as loss function, you can get the evaluation result, which includes AUROC, AUPR and LossLoss calculated by scikit-learn, following to below commands.

$ cd vw-luigi
$ ls /tmp/work/space
> train.vw test.vw
$ python -m luigi --module vwluigi EvalTask --loss-func squared --work-dir /tmp/work/space --local-scheduler
> ...
$ ls /tmp/work/space
> model.vw predict.vw result.txt train.vw
$ cat /tmp/work/space/result.txt
> AUROC:0.88060 AUPR:0.72192 LOGLOSS:0.36215

If you are interested in vw-luigi, please see this blog post "'Kaggle Display Advertising Challenge' working with vw-luigi". I wrote another usage example using 'Kaggle Display Advertising Challenge Dataset' provided by Critio.

• 0.1.0
  • The first proper release

Distributed under the MIT license. See LICENSE for more information.
Author: Shotaro Kohama - tw: @shotarok28