pytorch==1.0.0 numpy gensim nltk pyhocon argparse transformers

• DApredict.py: DA prediction or estimation model
• NLI.py: NLI model for reward
• run_glue.py: NLI model trainer
• order_predict.py: dialogue order predict model
• utils.py: utility tools
• models.py: response generation model
• nn_blocks.py: neural network components
• preprocess.py: preprocess dataset
• evaluation.py: evaluate response generation model
• quantitative_evaluation.py: calc. BLEU, Distinct, ...
• train.py: response generation model trainer
• experiments.conf: config file
• data/
  • corpus/ :corpus
  • model/ :trained model
  • result/ : results

1. train HRED model

python train.py --expr HRED_dd --gpu <gpu num>

2. reinforcement learning

python train.py --expr RL_dd --gpu <gpu num>

3. evaluation

python evaluation.py --expr RL_dd --gpu <gpu num>
python quantitative_evaluation --expr RL_dd

https://drive.google.com/open?id=1g42PHKWzmmzUaA8vK02b7ge9TL_z6ZfK