The process.py read all the tabular files from in the YYMMDD-batchname directory and generate one file per GUID to the golds directory
I found that most of the rows in the table were missing content, and only a few were actually labeled, so that's why if you randomly check one generated file in 'golds', the file content will not match what you imagine :)
The process.py read all the tabular files from in the YYMMDD-batchname directory and generate one file per GUID to the golds directory