twmkn9/bert-base-uncased-squad2

export SQUAD_DIR=../../squad2
python3 run_squad.py 
--model_type bert 
--model_name_or_path bert-base-uncased 
--do_train 
--do_eval 
--overwrite_cache 
--do_lower_case 
--version_2_with_negative 
--save_steps 100000 
--train_file $SQUAD_DIR/train-v2.0.json 
--predict_file $SQUAD_DIR/dev-v2.0.json 
--per_gpu_train_batch_size 8 
--num_train_epochs 3 
--learning_rate 3e-5 
--max_seq_length 384 
--doc_stride 128 
--output_dir ./tmp/bert_fine_tuned/