Int4/8 Quantized Whisper Models by using the quanto package and the CTranslate2 package. Requires (much) less GPU resources while keeping performance.