Please create one issue per bullet point when you start working on them and edit this message to add the link to it. - [ ] Use PyTorch and PyTorch Lightning - [ ] Switch to Precision/Recall/F1 (Optionally BLEU and ROUGE also) - [ ] Add SotA section, highlighting the results from publications and NLP domain (size of the dataset, etc) - [ ] Use Tree-sitter for preprocessing - [ ] Include more experiments (ablation study, effect of the dataset size, GGNN model, comparison of structured and non-structured features) - [ ] Make it simpler (move code to external file, use collapsible cells)