Attention-based Tree-to-Sequence Code Summarization Model

The TensorFlow Eager Execution implementation of Source Code Summarization with Extended Tree-LSTM (Shido+, 2019)

including:

• Multi-way Tree-LSTM model (Ours)
• Child-sum Tree-LSTM model
• N-ary Tree-LSTM model
• DeepCom (Hu et al.)
• CODE-NN (Iyer et al.)

Dataset

1. Download raw dataset from [https://github.com/xing-hu/DeepCom]
2. Parse them with parser.jar

Usage

1. Prepare tree-structured data with dataset.py
   • Run $ python dataset.py [dir]
2. Train and evaluate model with train.py
   • See $ python train.py -h