### 1. fine-tune CodeBERT on NL-Code Search Task

python code/run_classifier.py
			--model_type roberta
			--do_predict
			--test_file test_webquery.json
			--max_seq_length 200
			--per_gpu_eval_batch_size 2
			--data_dir data
			--output_dir ./model_codesearchnet/checkpoint-best-aver
			--encoder_name_or_path microsoft/codebert-base
			--pred_model_dir ./model_codesearchnet/checkpoint-last
			--prediction_file ./evaluator/webquery_predictions.txt 
			
### 2. attack

python code/run_attack.py 
			--data_dir ./data_codebert/ 
			--output_dir ./model_codesearchnet/checkpoint-best-aver 
			--encoder_name_or_path microsoft/codebert-base 
			--pred_model_dir ./model_codesearchnet/checkpoint-last/

cd Document_Translation

### 1. fine-tune on translation code documentation task

sh run-multi.sh

### 2. attack

python run_attack_from_json.py