diff --git a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala
index 45b9ebb4cc0d6..9fd60ff7a0c79 100644
--- a/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala
+++ b/mllib/src/main/scala/org/apache/spark/mllib/regression/GeneralizedLinearAlgorithm.scala
@@ -211,6 +211,10 @@ abstract class GeneralizedLinearAlgorithm[M <: GeneralizedLinearModel]
    */
   def run(input: RDD[LabeledPoint], initialWeights: Vector): M = {
 
+    if (numFeatures < 0) {
+      numFeatures = input.map(_.features.size).first()
+    }
+
     if (input.getStorageLevel == StorageLevel.NONE) {
       logWarning("The input data is not directly cached, which may hurt performance if its"
         + " parent RDDs are also uncached.")
diff --git a/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala
index aaa81da9e273c..a26c52852c4d7 100644
--- a/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/mllib/classification/LogisticRegressionSuite.scala
@@ -425,6 +425,12 @@ class LogisticRegressionSuite extends FunSuite with MLlibTestSparkContext with M
 
     val model = lr.run(testRDD)
 
+    val numFeatures = testRDD.map(_.features.size).first()
+    val initialWeights = Vectors.dense(new Array[Double]((numFeatures + 1) * 2))
+    val model2 = lr.run(testRDD, initialWeights)
+
+    LogisticRegressionSuite.checkModelsEqual(model, model2)
+
     /**
      * The following is the instruction to reproduce the model using R's glmnet package.
      *