this is a QASystem implemented with BERT
install with command
pip3 install -U tf-nightly-2.0-preview bert-for-tf2 flask flask-cors flask-socketio celery gevent
sudo apt install libboost-all-dev rabbitmq-server
celery and socketio need rabbit message queue, so launch it with command
sudo systemctl start rabbitmq-server
download with the following command.
bash downloads.sh
put the questions and answers in format as question_answer.txt's. and execute following command to convert the collected samples into dataset format.
make -C cc && make -C cc install
./cc/create_dataset -i question_answer.txt -o dataset
with directory dataset generated by the above command presented, finetune with the following to start finetune.
python3 Predictor.py
run the QA system server by
CUDA_VISIBLE_DEVICES='' python3 server.py
stop the server by Ctrl+C
firefox <ip>:5000