text_simplification
- ### Arugument instruction - bsize: batch size - out: the output folder will contains log, best model and result report - tie_embedding: all means tie the encoder/decoder/projection w embedding, we found it can speed up the training - bert_mode: the mode of using BERT bert_token indicates we use the subtoken vocabulary from BERT; bertbase indicates we use BERT base version (due to the memory issue, we did not try BERT large version yet) - environment: the path config of the experiment. Please change it in model/model_config.py to fit to your system