languages=tr_only
vocab_size=8000
// TuncerLab Cortex Tokenizer