Fully open-sourced streaming speech-to-text and speech-to-speech translation model with state-of-the-art performance