# Progressive model compression

This is a multilingual Roberta model that can generate embeddings for Bulgarian sentences. Trained based on Sentence-BERT principles with Google's USE model as the teacher model.