Bug in loading logic

[mgdenno]

From Teams message exchange:

so we first convert the parquet files to the the required format and save to the cache dir

but when we read from the cache dir we read all the files in the cache dir

but we don't clear the cache dir each time

so the cache dir currently has both the good file in it but also the old failed files

so it is trying to load the failed ones too...that is the issue

...so we either need to clear the cache before running or use explicit file names...