python run_qa.py \
  --model_name_or_path bert-base-uncased \
  --dataset_name squad \
  --do_train \
  --do_eval \
  --per_device_train_batch_size 12 \
  --learning_rate 3e-5 \
  --num_train_epochs 2 \
  --max_seq_length 384 \
  --doc_stride 128 \

***** eval metrics *****
  epoch                   =        2.0
  eval_exact_match        =    81.3434
  eval_f1                 =    88.7002
  eval_samples            =      10784