The Tanakh dataset can be used for training and fine-tuning Biblical Hebrew language models. It consists of the vocalized text of the Hebrew Bible (excluding the Aramaic verses in Jeremiah 10:11; Daniel 2:4b–7:28; Ezra 4:8–6:18; 7:12-26) broken down by verse. The base text of each verse comes from the Text-Fabric package and was subsequently processed in order remove accents and other paratextual markers as well as perform custom Unicode normalization using hebrewtools.
The script used to generate the Tanakh dataset can be found here.