Python DataFrame type inference for LabeledPoint gets wrong type

In Pyspark, when an RDD of LabeledPoints is converted to a DataFrame using toDF(), the returned DataFrame has type "null" instead of VectorUDT.

To reproduce:

from pyspark.mllib.util import MLUtils
rdd = MLUtils.loadLibSVMFile(sc, "data/mllib/sample_libsvm_data.txt")
df = rdd.toDF()

Examine rdd and df to see:

>>> df
DataFrame[features: null, label: double]