chunked/nltk_chunking/tasks_multiple_choice/chunk_14.txt

You can speed up the map function by setting batched=True to process multiple elements of the dataset at once:
py
tokenized_swag = swag.map(preprocess_function, batched=True)
🤗 Transformers doesn't have a data collator for multiple choice, so you'll need to adapt the [DataCollatorWithPadding] to create a batch of examples.