1

I get the error below when I select to run XGBOOST via the menu in H2O-3 flow UI or when I include it when running all architectures (also via the "AutoML" options in H2O-3 flow). I only select the train + validation frames + response column and the click "build model".

Btw: I have set the following 2 enviroment variables:

Name: JRE_HOME + JAVA_HOME
Value: C:\Program Files\Java\jre1.8.0_211

I'm running Windows 10

java.lang.UnsatisfiedLinkError: ml.dmlc.xgboost4j.java.XGBoostJNI.XGDMatrixCreateFromMat([FIIF[J)I

10-24 21:05:31.201 192.168.0.126:54321 30544 FJ-1-7 INFO: Rebalancing train dataset into 8 chunks.
10-24 21:05:31.208 192.168.0.126:54321 30544 FJ-1-7 INFO: Completing model xgboost-ca6b679e-bed4-41a0-9129-9a28f38c8cc5
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: java.lang.UnsatisfiedLinkError: ml.dmlc.xgboost4j.java.XGBoostJNI.XGDMatrixCreateFromMat([FIIF[J)I
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at ml.dmlc.xgboost4j.java.XGBoostJNI.XGDMatrixCreateFromMat(Native Method)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at ml.dmlc.xgboost4j.java.DMatrix.(DMatrix.java:183)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.tree.xgboost.XGBoost.hasGPU_impl(XGBoost.java:600)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.tree.xgboost.XGBoost.hasGPU(XGBoost.java:581)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.tree.xgboost.XGBoost.access$000(XGBoost.java:32)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.tree.xgboost.XGBoost$XGBoostDriver.buildModel(XGBoost.java:242)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.tree.xgboost.XGBoost$XGBoostDriver.computeImpl(XGBoost.java:237)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at hex.ModelBuilder$Driver.compute2(ModelBuilder.java:222)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at water.H2O$H2OCountedCompleter.compute(H2O.java:1417)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at jsr166y.CountedCompleter.exec(CountedCompleter.java:468)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at jsr166y.ForkJoinTask.doExec(ForkJoinTask.java:263)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at jsr166y.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:974)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at jsr166y.ForkJoinPool.runWorker(ForkJoinPool.java:1477)
10-24 21:05:31.209 192.168.0.126:54321 30544 FJ-1-7 ERRR: at jsr166y.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:104)
ashwin agrawal
  • 1,603
  • 8
  • 16
PabloDK
  • 2,181
  • 2
  • 19
  • 42

1 Answers1

2

I guess you are facing this error because:-

H2o xgboost is not currently supported in Windows.

enter image description here

ashwin agrawal
  • 1,603
  • 8
  • 16
  • I see...the wierd part is that XGBoost does support Windows (I can run it separately in Windows via the CMD)...And after I build the XGBoost dll and copied it to my java lib folder H2O also loads Xgboost when it starts up + makes it visible in the UI...so everything seems to work in regards to installation/setup - but eventually it doesn't when I actually try to use it :( – PabloDK Oct 28 '19 at 13:14
  • Did you run the `xgboost` on vanilla windows, or did you setup some virtual env to run your model. – ashwin agrawal Oct 28 '19 at 15:41
  • plain vanilla Windows 10 64 bit – PabloDK Oct 28 '19 at 16:55
  • I will have to check then, as still `xgboost`, is not available for windows. – ashwin agrawal Oct 28 '19 at 16:58
  • 1
    FYI (I ended up installing Ubuntu as my second OS): XGBOOST is integrated into the Linux version (you don't have to manually compile the dll first - like in the Windows version). – PabloDK Nov 12 '19 at 06:08
  • Also, I made H2O eat my compiled XGBOOST dll file (I placed it in the Java/lib dir and restarted the H2O engine). But even though H2O stated that it now includes and loads XGBOOST as well - it eventually didn't work out...It throw some kind of exception when I tried to use it...and I couldn't find/google a solution to it... – PabloDK Nov 12 '19 at 10:52