JAX implementation of OpenAI's Whisper model for up to 70x speed-up on TPU. - View it on GitHub