jkot/transormer_en_cs_translator

To train model and run evaluation:

Download and extract the europarl cs_en from https://www.statmt.org/europarl/ to datasets/europarl/ folder

Install requirements using pip install -r requirements.txt

Run scripts in order create_dataset_splits.py, train_tokenizers.py, preprocess_dataset.py, train.py and eval.py

In the models folder, there is already trained model in .keras format and there are also dataset splits and pretrained tokenizers, so you can run evaluation directly