Can i reconstruct .wav file from jukebox embeddings?

hello, i try to extract the embedding with: def extract(fpath, skip_completed=True, dest_dir="aist_juke_feats"): os.makedirs(dest_dir, exist_ok=True) audio_name = Path(fpath).stem save_path = os.path.join(dest_dir, audio_name + ".npy")

if os.path.exists(save_path) and skip_completed: return

audio = jukemirlib.load_audio(fpath) # default sample rate of Jukebox is 44100 reps = jukemirlib.extract(audio, layers=[LAYER], downsample_target_rate=FPS)

#np.save(save_path, reps[LAYER]) return reps[LAYER], save_path

now i wanna reconstruct audio .wav from reps[66], but as far as i know, jukebox vqvae decode need "vqvae code", but i only have "reps[66]", what should i do?