sort_metric argument in h2o.automl does not get used

Question

I'm trying to run an h2o automl model for a multi class problem in the following manner

h2o.automl(y = result,
              training_frame = train,
              max_runtime_secs = 30,
              sort_metric = "logloss")

But I end up getting an error

Error in h2o.automl(y = result, training_frame = train_to_model, max_runtime_secs = 30, : unused argument (sort_metric = "logloss")

Based on the documentation in http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/algo-params/sort_metric.html I should be able to use logloss in a multi class classification problem but I'm not able to.

My h2o version is 3.18.0.11 and R version is 3.4.1

Can you please tell me why this might be happening? Thank you!

score 1 · Answer 1 · answered Jun 13 '18 at 16:20

The sort_metric argument was added in the most recent stable release of H2O, 3.20.0.1. You must upgrade to the latest version of H2O to use it. You can always tell which version of H2O is the latest version by looking at the version number listed in the user guide. See screenshot below.

sort_metric argument in h2o.automl does not get used

1 Answers1