Spark does not read files that are hidden. The condition for the hidden file is if the files starts with underscore(_) or period . (for Hadoop _ indicates hidden as per the comments in https://issues.apache.org/jira/browse/SPARK-26339)
I am assuming since the checkpoint file starts with _ it is not being read. One solution I can think of is copying the file to another directory with a different name and then reading it from spark.