Threshold metrics calculation fix when unseen labels are present

core/src/main/scala/com/salesforce/op/evaluators/OpMultiClassificationEvaluator.scala

@@ -188,7 +188,8 @@ private[op] class OpMultiClassificationEvaluator
     def computeMetrics(scoresAndLabels: (Array[Double], Double)): MetricsMap = {
       val scores: Array[Double] = scoresAndLabels._1
       val label: Label = scoresAndLabels._2.toInt
-      val trueClassScore: Double = scores(label)
+      // The label may be unseen during model training, so treat scores for unseen classes as all being zero
+      val trueClassScore: Double = scores.lift(label).getOrElse(0.0)
       val topNsAndScores: Map[Label, Array[(Double, Int)]] = topNs.map(t => t -> scores.zipWithIndex.sortBy(-_._1)
         .take(t)).toMap
       val topNScores: Map[Label, Array[Double]] = topNsAndScores.mapValues(_.map(_._1))

core/src/test/scala/com/salesforce/op/stages/impl/classification/MultiClassificationModelSelectorTest.scala

@@ -405,6 +405,16 @@ class MultiClassificationModelSelectorTest extends FlatSpec with TestSparkContex
             .setInput(label, features)
         testEstimator.fit(bigDataWithMeta)
 
+        val prediction = testEstimator.getOutput()
+        val model = testEstimator.fit(bigDataWithMeta)
+        val transformedBigData = model.transform(bigDataWithMeta)
+
+        // Verify that multiclass metrics can be computed even when unseen labels may be present
+        val evaluatorMulti = new OpMultiClassificationEvaluator()
+          .setLabelCol(label)
+          .setPredictionCol(prediction)
+        noException should be thrownBy evaluatorMulti.evaluateAll(transformedBigData)
+
         val (numLabelsInCutter, numLabelsInCutterMetadata) = {
           import scala.language.reflectiveCalls
           val df = cutter.prevalidationValForTest