2

I'm trying to run a h2o.automl script, nothing too complex, that has worked before plenty of times on my current machine

autoML= h2o.automl(x = x,
               y = y, 
               training_frame = data.trainingtest.h2o, 
               leaderboard_frame = data.validation.h2o, 
               max_runtime_secs = 3600*3,
               exclude_algos = c("StackedEnsemble", "DeepLearning"))

The problem is that I get the following error, and I don't know where to start looking for a solution

Error in .h2o.doSafeREST(h2oRestApiVersion = h2oRestApiVersion, urlSuffix = page, : Unexpected CURL error: getaddrinfo() thread failed to start

The problem arises when it is almost finished, so it is really annoying because the first times I ran it for three hours. Is there something that I can add to make the problem more approachable? has anyone faced something like this?

UseR10085
  • 7,120
  • 3
  • 24
  • 54
Gabriel Gajardo
  • 65
  • 7
  • I run into the same issue when trying to run in v3.32.0.3 h2o and R v4.0.3. It seems that if you try to run multiple h2o autoML in a loop (I did 100), you can run into this issue around the 10th autoML. – Chris Toh Jan 21 '21 at 15:55
  • I got the same error running R 4.0.3 on Ubuntu 20.04, but only when the response y is multinomial. Models running binomial and numeric responses are working fine with `h2o.automl`. I'm using a brand new installation of Ubuntu 20.04, and installed Java `openjdk-8-jre-headless` and `jdk-11.0.10_linux-x64`. The [h2o documentation](http://docs.h2o.ai/h2o/latest-stable/h2o-docs/welcome.html) indicates the supported versions include Java 8, 9, 10, 11, 12, 13 and 14, but I just don't know what is the best JDK/JDE combination, if it exists. – zabala Jan 29 '21 at 17:24

0 Answers0