Hello,
Okay, I found why my toy worked. I call
AbstractEventTrainer.doTrain(DataIndexer) as oppose to
AbstractEventTrainer.train(ObjectStream<Event>). The train method calls
isValid(). That sets the value of threads in QNTrainer.
Thank you for making me do this. I don’t think doTrain() should be exposed
anymore. A new method train(DataIndexer) that calls isValid and then
doTrain(indexer) is probably a better idea. Is it important to calculate the
hash of all events?
Daniel
On 10/27/16, 11:49 AM, "Russ, Daniel (NIH/CIT) [E]" <dr...@mail.nih.gov> wrote:
Comment 1:
Here is my Stacktrace:
Exception in thread "main" java.lang.IllegalArgumentException: Number of
threads must 1 or larger
at
opennlp.tools.ml.maxent.quasinewton.ParallelNegLogLikelihood.<init>(ParallelNegLogLikelihood.java:50)
at
opennlp.tools.ml.maxent.quasinewton.QNTrainer.trainModel(QNTrainer.java:165)
at
opennlp.tools.ml.maxent.quasinewton.QNTrainer.doTrain(QNTrainer.java:152)
at
opennlp.tools.ml.maxent.quasinewton.QNTrainer.doTrain(QNTrainer.java:1)
at gov.nih.cit.test.smlm.SOCcerME.train(SOCcerME.java:131)
at gov.nih.cit.test.smlm.SOCcerME.main(SOCcerME.java:310)
So I threw together a small toy to get the same error. It worked… Let me
track this down some more.
