H2O AutoML训练时触发NullPointerException问题咨询

Environment #

• H2O Version: Stable release 3.30.1.3
• R Version: 4.0.2
• Operating System: macOS Catalina 10.15.6
• Dataset: ~20,000 observations (sensitive data, no reproducible example available)

Problem #

I’ve been working with H2O for roughly 3 years and this is the first time I’ve hit this specific error. During an AutoML training run, the process crashed at 53% completion with a NullPointerException. The full stack trace is included below.

Error Stack Trace #

java.lang.NullPointerException
java.lang.NullPointerException
at hex.ModelMetrics.getMetricFromModelMetric(ModelMetrics.java:151)
at ai.h2o.automl.leaderboard.Leaderboard.getMetrics(Leaderboard.java:558)
at ai.h2o.automl.leaderboard.Leaderboard.updateModels(Leaderboard.java:422)
at ai.h2o.automl.leaderboard.Leaderboard.lambda$addModels$0(Leaderboard.java:381)
at ai.h2o.automl.leaderboard.Leaderboard.atomicUpdate(Leaderboard.java:442)
at ai.h2o.automl.leaderboard.Leaderboard.addModels(Leaderboard.java:378)
at ai.h2o.automl.leaderboard.Leaderboard.addModel(Leaderboard.java:459)
at ai.h2o.automl.ModelingStepsExecutor.addModel(ModelingStepsExecutor.java:186)
at ai.h2o.automl.ModelingStepsExecutor.monitor(ModelingStepsExecutor.java:163)
at ai.h2o.automl.ModelingStepsExecutor.submit(ModelingStepsExecutor.java:82)
at ai.h2o.automl.AutoML.learn(AutoML.java:604)
at ai.h2o.automl.AutoML.run(AutoML.java:407)
at ai.h2o.automl.H2OJob$1.compute2(H2OJob.java:33)
at water.H2O$H2OCountedCompleter.compute(H2O.java:1563)
at jsr166y.CountedCompleter.exec(CountedCompleter.java:468)
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)

My Hypothesis #

I’m wondering if this issue might stem from using the same dataset for both the training_frame and leaderboard_frame parameters. If there’s any additional information I can provide to help diagnose this, please let me know—I appreciate any guidance!

内容的提问来源于stack exchange，提问作者Bernardo