test_train_eval() occasionally fails because the test F1 remains zero. Now, we ensure reproducibility by setting the seed and trainer.deterministic=True. Furthermore, we just check that the test scores are the same as after training (and do not ensure that it is above zero).
test_train_eval()
occasionally fails because the test F1 remains zero. Now, we ensure reproducibility by setting theseed
andtrainer.deterministic=True
. Furthermore, we just check that the test scores are the same as after training (and do not ensure that it is above zero).