Question-Answer special tokens

model/supervised_finetuning/custom_datasets/__init__.py

@@ -2,6 +2,12 @@ from datasets import load_dataset
 from sklearn.model_selection import train_test_split
 from torch.utils.data import Dataset, Subset
 
+QA_SPECIAL_TOKENS = {
+    'Question': '<question>', 
+    'Answer': '<answer>'
+}
+
+
 
 class SquadV2Dataset(Dataset):
     def __init__(self, cache_dir, split):