Array index out of range: 9446 - Error During the Creation of Deep Learning Model

[Christopher Marin]

I've been playing around a bit with H2) and I have to say I'm very impressed.  The first time I tried it I was able to create some Deep Learning and Random Forest models with ease within the web UI.  I did run into an error though today when I tried to create a Deep Learning model. I got this exception:

DeepLearningProgressPage

error

Got exception 'class java.lang.ArrayIndexOutOfBoundsException', with msg 'Array index out of range: 9446'

java.lang.ArrayIndexOutOfBoundsException: Array index out of range: 9446

at java.util.Arrays.rangeCheck(Arrays.java:1311)

at java.util.Arrays.binarySearch(Arrays.java:1454)

at water.fvec.NewChunk.values(NewChunk.java:84)

at water.fvec.RebalanceDataSet$RebalanceTask.rebalanceChunk(RebalanceDataSet.java:117)

at water.fvec.RebalanceDataSet$RebalanceTask.map(RebalanceDataSet.java:139)

at water.MRTask2.compute2(MRTask2.java:404)

at water.MRTask2.compute2(MRTask2.java:365)

at water.MRTask2.compute2(MRTask2.java:365)

at water.MRTask2.compute2(MRTask2.java:365)

at water.MRTask2.compute2(MRTask2.java:365)

at water.H2O$H2OCountedCompleter.compute(H2O.java:634)

at jsr166y.CountedCompleter.exec(CountedCompleter.java:429)

at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263)

at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:974)

at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477)

at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104)

I was able to create a Random Forest model using the same dataset.  I tried splitting the data, severely restricting the number of columns analyzed, checking and unchecking the rebalance dataset option, and restarting the server with more memory.  Nothing seemed to make a difference, and I kept getting that error. Any ideas?

[arno....@gmail.com]

Hi Chris,

Thanks for reporting that problem. Is there a chance you can share the log files (and a dataset that reproduces this problem)? This has been an elusive bug so far: https://0xdata.atlassian.net/browse/PUB-1012

Thank you!
Arno

[arno....@gmail.com]

Admin->Inspect Logs->Download All Logs

You can also respond to sup...@0xdata.com if you can't share your data publicly.

Thanks,
Arno

[Christopher Marin]

Much appreciated Arno!  I replied offline to that address.