git clone https://gitlab.mff.cuni.cz/semind/hw1-diacritics-restoration
pytorch
unidecode
conda env create --name semind_hw1 --file=environment.yml
conda activate semind_hw1
conda deactivate semind_hw1
conda env remove --name semind_hw1
use make all
to download the data, train and evaluate on the dev dataset
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
use make clean
to erase trained weights and data
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