loodos/bert-base-turkish-uncased

Turkish BERT-Base (uncased) model has 12 encoder layers with 768 hidden layer size trained on uncased Turkish dataset . Huggingface's tokenizers (which were written in Python) have a bug concerning letters "ı, i, I, İ" and non-ASCII Turkish specific letters . Until it is fixed, we provide a simple text normalization module (TextNormalization() in the code snippet above) to train your model with uncased data . The default string.lower() and string.upper() make the conversions "I" and "İ" to 'i' and 'ı' to 'I' in Turkish, but in Turkish 'I,