The code is based on PyTorch 1.7.0 and Python 3.7.7. For training, a GPU is recommended to accelerate the training speed.

The code is based on Python 3.7. Its dependencies are summarized in the file requirements.txt.

numpy==1.20.1
torch==1.7.0
tqdm==4.62.3
transformers==4.19.2

You can install these dependencies like this:

pip3 install -r requirements.txt

• Pretrain the full model on Conll2003 dataset first, then train it on specific target domain with default hyperparameter settings . Use target domain as an example.    

      python3 main.py --num_source_tag 17 --batch_size 16 --tgt_dm science

        num_source_tag should equal to the number of classes of source domain, num_target_tag should equal to the number of classes of target domain and --src_dm is the source domain name, --tgt_dm is the target domain name.

You need to declare all your domain labels at first in ./src/dataloader.py line 29. Each dataset is a folder under the ./ner_data folder. You must name your folder as its domain name, which must be the same as the --src_dm, --tgt_dm parameter:

./ner_data
└── source
    ├── train
    ├── dev
    ├── test