The gold annotation for the TDT official test set is not publicly distributed. Instead, you can submit your parser output on the file fi_ud_test-official.conllu to the evaluation service at the address below. The service will compare your parser output against the gold standard and will give you the scores for a number of standard metrics. These are to be considered the "official" test set results. The files fi_ud_devel.conllu and fi_ud_test.conllu are obtained by splitting the original development set of TDT into two equally-sized parts. TDT evaluation service address: http://bionlp-www.utu.fi/tdteval/