Judging Adam: Studying the Performance of Optimization Methods on ML4SE Tasks

This is an official repository containing the experiments reported in

Pasechnyuk D.A., Prazdnichnykh A., Evtikhiev M., Bryksin T. "Judging Adam: Studying the Performance of Optimization Methods on ML4SE Tasks",

which is submitted to ICSE NIER'2023.

Source code for Code2Seq, TreeLSTM and CodeTransformer models, which were used in course of our research, was adapted by us based on existing implementations. The list of corresponding original repositories follows:

• Code2Seq by SpirinEgor
• TreeLSTM by SpirinEgor
• CodeTransformer by danielzuegner

The CodeGNN model's implementation with PyTorch is original, it's based on the Tensorflow implementation from

• CodeGNN by acleclair

For running training and validation, follow the instructions in corresponding README.md files in code2seq, treelstm, codetransformer and codegnn.