Repository for the bachelor's thesis:
Evaluation Of Automated Code Documentation Approaches
The preprocessing procedure can be invoked by using preprocess_main.py. This file triggers all preprocessing procedures sequentially.
The training has to be invoked seperately for each approach. Details on how to train each models are available at the authors repositories: attnFc, code2seq, CodeGNN, NeuralLSPS
Some training procedures require additional files than can be easily found at the author's repository for download.
All approaches can be evaluated sequentially with evaluate_main.py. This requires the preprocessed data and the prediction files from the models.
Link to download final model files for each approach: HERE
Note: Some hardcoded paths require adjustment to fit a different directory structure