removed dataset caching as it seems to impact too much on the memory footprint

dhp-workflows/dhp-graph-mapper/src/main/resources/eu/dnetlib/dhp/oa/graph/raw_all/oozie_app/workflow.xml

@@ -639,6 +639,7 @@
                 --conf spark.sql.queryExecutionListeners=${spark2SqlQueryExecutionListeners}
                 --conf spark.yarn.historyServer.address=${spark2YarnHistoryServerAddress}
                 --conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
+                --conf spark.sql.shuffle.partitions=8000
             </spark-opts>
             <arg>--master</arg><arg>yarn</arg>
             <arg>--hdfsPath</arg><arg>${workingDir}/graph_raw</arg>

dhp-workflows/dhp-graph-mapper/src/main/scala/eu/dnetlib/dhp/oa/graph/raw/CopyHdfsOafSparkApplication.scala

@@ -62,7 +62,6 @@ object CopyHdfsOafSparkApplication {
       val oaf = spark.read
         .textFile(validPaths: _*)
         .map(v => (getOafType(v), v))(Encoders.tuple(Encoders.STRING, Encoders.STRING))
-        .cache()
 
       try {
         ModelSupport.oafTypes