drop other labels like 'answer' and make label int

get_all.py

@@ -61,6 +61,18 @@ datasets = DatasetDict({
 
 datasets = datasets.filter(drop_na)
 
+labels = {
+    'entailment': 2,
+    'neutral': 1,
+    'contradiction': 0,
+}
+def label_map(row):
+    row['labell'] = list(map(lambda x: labels[x], row['label']))
+    return row
+datasets = datasets.filter(lambda row: row['label'] in labels.keys())
+datasets = datasets.map(label_map, batched=True, remove_columns=['label'])
+datasets = datasets.rename_column('labell', 'label')
+
 datasets.push_to_hub("seongs1024/DKK-nli", private=True)
 
 # datasets = load_dataset('seongs1024/DKK-nli')