# Fill-mask

A pre-trained Amharic language model based on the bert-medium architecture, with 40.5 million parameters trained on 290 million tokens, achieving performance comparable to larger multilingual models.