Pytorch Template for Named Entity Recognition (BERT)

Overview

This repository contains Pytorch implement of named entitu recognition. You can define parameters in main.py, including batch_size, learning_rate, n_epochs etc. The parameters will save in work_dir. You can also use custom config file by following command.

python main.py -c [your config file address]

Argument description

work_dir -> Directory to save args.
model_name -> Pretrained model name you want to use in embedding layer
logdir -> Directory to save target v.s. prediction
train_type -> Different model for NER, including PLM-BiLSTM-CRF, PLM-CRF
patience -> Eealystopping patience for dev dataset

Environment

• numpy==1.23.2
• scikit_learn==1.1.2
• torch==1.12.1
• tqdm==4.64.0
• transformers==4.18.0
• yaml~=0.2.5
• pyyaml~=6.0
• scikit-learn~=1.1.2
• pandas~=1.2.5