BEReshiT model

This repository contains: the Ancient Hebrew model BEReshiT, which has been achieved through the finetuning of layers 9-12 of the Modern Hebrew model DictaBERT config.json model.safetensors special_tokens_map.json tokenizer.json tokenizer_config.json vocab.txt the morphological submodel BEReshiT-morph, which predicts morphological labels relevant to Ancient Hebrew text best_hparams config_morph head_weights label2id pytorch_model.bin special_tokens_map_morph tokenizer_config_morph tokenizer_morph vocab_morph related train and evaluation dataset all_mlm.csv - used for model training eval_200_genre_and_grammar_with_masks.csv - evaluation dataset divided into the categories "by category" and "by phenomenon" dss_lines_mlm.csv - evaluation dataset based on the Dead Sea Scrolls)