Luna_yyq
Results
1
comments of
Luna_yyq
I use lm-evaluation-harness v0.4.0 for evaluation, which is consistent with the evaluation version used by the author. In addition, the results displayed above are obtained using num_train_epochs=6 for training.