nusnlp / esc Goto Github PK

Trying to reproduce the paper results crashes the script.
Running:export EXP_DIR=conll-exp
and then: python run.py --test --data_dir $EXP_DIR/test-text --m2_dir $EXP_DIR/test-m2 --model_path $EXP_DIR/models/paper_model.pt --vocab_path $EXP_DIR/paper_vocab.idx --output_path $EXP_DIR/outputs/test.out

yields the following error:

Traceback (most recent call last):
  File "run.py", line 542, in <module>
    main(args)
  File "run.py", line 516, in main
    with open(args.output_path, 'w', encoding='utf-8') as out:
FileNotFoundError: [Errno 2] No such file or directory: 'conll-exp/outputs/test.out'

reproducing the bea-2019 experiments and re-training works fine.

Hi im trying to use this github repo to evaluate GEC tasks on other languages other than English. I have the (.m2) files but some languages dont have tools like ERRANT. In that case is it still possible to run the ensemble method? if so what are the necessary files?

Thank you!!

Hello,
I have two questions,

1. I saw dev-text folder and it had more than one .txt file. I know about source/target file ( it has incorect/correct pair , right?) but what about other files that has same sentences with different order/grammar?

2. From what I have found, for training phase we need source/target file (incorrect/correct sentence) but what about M2 format? Is it for just enhancing the model or other things?