edge-prediction-based-CQA-model

"# edge-prediction-based-CQA-model" Directory structure: layers: encoder: the encoding layer. Input: paragraph, conversation history, current question. output: paragraph, conversation history, current question with embeddings of each words reasoner: the reasoning layer. Input: the output of encoder layer output: evidence with embeddings of each words predicter: the prediction layer. Input: the output of reasoner layer output: answer (answer type, start position, end position) interface: define the output of each layer models: the models class we use like BERT and our proposed model. proposed_model.py will call functions of layer.py trainer.py: do the training, evaluation, testing in the minibatch. call proposed_model.py main.py: read args from cmd and start the process call the trainer.py utils: the common utils used by multi layers or the whole model config.py: mutable configuration like epochs=5, output_dir data: data log: log model: trained model parameters. save them here. requirement.txt: required packages