The codes for testing Ext-MNLI&SNLI&SICK/Simple-Premise/Simple-Pair by BERT/RoBERTa/DeBERTa
0. DATASETS file includes Simple Pair (N-is-A.csv/SVO.csv/N-is-A_conjun.csv/SVO_conjun.csv/N-is-A_coreference.csv/SVO_coreference.csv), Extended Pair (extended_simple.csv/extended_conjun.csv), and augmented training set (aug_train.csv)
1. Install all packages in requirement.txt.
2. Run training.py to fine-tune BERT/ALBERT/RoBERTa models or augmented training. The parameters for fine-tuning are shown in Table.
3. Run the codes (SVO*.py/N_is_A*.py/Extended*.py) to test models on the 3 datasets
4. Run the code (test_dev.py) to test models on the corresponding test datasets.

