Removed relativeSD validation from python API,RDD.scala will do valid…

…ation

python/pyspark/rdd.py

@@ -2237,10 +2237,7 @@ def countApproxDistinct(self, relativeSD=0.05):
         >>> 16 < n < 24
         True
         """
-        if relativeSD < 0.000017:
-            raise ValueError("relativeSD should be greater than 0.000017")
-        if relativeSD > 0.37:
-            raise ValueError("relativeSD should be smaller than 0.37")
+
         # the hash space in Java is 2^32
         hashRDD = self.map(lambda x: portable_hash(x) & 0xFFFFFFFF)
         return hashRDD._to_java_object_rdd().countApproxDistinct(relativeSD)