This is the repository for the demo backend of CoNLL 2020 Paper "What Are You Trying Todo? Semantic Typing of Event Processes". The opensource git repository (not including the demo backend) is https://github.com/CogComp/Event_Process_Typing
Environment Setup:
git clone https://github.com/CogComp/EventProcessTyping.git
cd EventProcessTyping
conda create --prefix ./STEP_env python=3.7
conda activate ./STEP_env
pip install -r requirements.txt
python -m spacy download en_core_web_sm
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
First, our running demo is here: http://dickens.seas.upenn.edu:4035/
Link to the pre-trained full models for console demo: https://drive.google.com/drive/folders/1b8peVVRNANL7r_Wnyyt4pPsyNROIlOfT?usp=sharing
After putting the file named
full_model_sptoken_ep121_a1.0_m1-0.1_m2-0.1.bin
under folder ./run_joint/full_model/
you can run the backend as follows:
cd run_joint/
CUDA_VISIBLE_DEVICES=0 nohup python backend.py 0 > nohup.out 2>&1 &
After running backend.py, you can check whether backend works with:
curl -d '{"sequence" : "dig a hole@put some seeds@fill the soil@water the soil"}' -H "Content-Type: application/json" -X POST http://dickens.seas.upenn.edu:4035/annotate/
Open url http://dickens.seas.upenn.edu:4035/ in the browser, then you can play with it!