--per_device_train_batch_size 2 --learning_rate 3e-5 --num_train_epochs 2 --max_seq_length 512 --doc_stride 250 --max_answer_length 200

--per_device_train_batch_size 2 --learning_rate 3e-5 --num_train_epochs 2 --max_seq_length 512 --doc_stride 250 --max_answer_length 200

initial commit