Playground for some experiments with Transformer models.
This folder contains code, which implements a transformer from scratch and trains it on the original Attention is all you need translation task. See the README in this folder for further details.
This folder contains the code to define and train my own BERT-like model DitBERT, which is also benchmarked with GLUE against BERT-base and RoBERTa-base. See the README in this folder for further details.
Contains the code used to parse the text corpora, used to train e.g. DitBERT.