Implement with tensorflow==1.14.
I take Question Generation
task as an example to show how model works.
The data is from this paper
You can run:
python preprocess.py
to generate data
You can run:
python run.py \
--model copynet \
--do_train \
--epoch 30 \
--batch 32 \
--optimizer Adam \
--lr 0.001 \
--dropout 0.2 \
--pre_train_epochs 5 \
--early_stop 5 \
--early_stop_delta 0.001