/NER

Pytorch-lightning implementation of Deep Transition RNNs

Primary LanguageJupyter Notebook

Named Entity Recognition

Pytorch-lightning implementation of the paper GCDT: A Global Context Enhanced Deep Transition Architecture for Sequence Labeling

drawing drawing
Rewards for grids

Installation

Training Environment

conda create --name train_ner python=3.8
conda activate train_ner
pip install -r requirements.txt

Inference Environment

conda create --name infer_ner python=3.8
conda activate infer_ner
pip install -r app/requirements.txt

Running the code

Training

This stores model checkpoints at lightning_logs/backup after every epoch

conda activate train_ner
python train.py --base_dir data/conll03 \
    --config_file configs/config.json \
    --epochs 1000

Test

Uses the files in lightning_logs/backup to output metrics as a csv file at data/conll03/testa_beam8.csv.

conda activate train_ner
python test.py

Infer on the webapp

conda activate infer_ner
python app/main.py