Fix seq2seq QA example

examples/pytorch/question-answering/run_seq2seq_qa.py

@@ -346,6 +346,7 @@ def main():
         raw_datasets = load_dataset(
             extension,
             data_files=data_files,
+            field="data",
             cache_dir=model_args.cache_dir,
             use_auth_token=True if model_args.use_auth_token else None,
         )