Install

git clone https://gitlab.mff.cuni.cz/semind/hw1-diacritics-restoration

Dependencies

pytorch
unidecode

Environment

Create

conda env create --name semind_hw1 --file=environment.yml

Activate

conda activate semind_hw1

Deactivate

conda deactivate semind_hw1

Remove

conda env remove --name semind_hw1

Usage

Download the data

use make all to download the data, train and evaluate on the dev dataset

Evaluation and print

use make eval to evaluate on the test dataset
use make print_dev to predict the text from the dev dataset and print it to stdout
use make print_test to predict the text from the test dataset and print it to stdout

Clean

use make clean to erase trained weights and data

Text to text script

use app.py to read from stdin and print the prediction to stdout

python app.py
tento text je nejlepsi
*ctrl-d or ctrl-z (windows) to stop the input*

also cat my_text_file.txt | app.py or for instance app.py < my_text_file.txt work