python version: 3.5.2
Pytorch version: 1.4.0
numpy version: 1.18.2
sklearn verison: 0.22.2.post1
tokenizers version: 0.5.2
transformers version: 2.5.1
nltk version: 3.4.5
spacy version: 2.2.4
gensim version: 3.8.1
other required packages:
preprocessor: https://pypi.org/project/tweet-preprocessor/
experiments were run on the NVIDIA GeForce GTX 1080 Ti GPUs.

command to run the code:
python3 vae_lbsoinn.py --train_path TRAIN_PATH --dev_path DEV_PATH --test_path TEST_PATH --cand_limit 1000 --keep_all_node --fix_per_task_mem_size
