LL-VQ-VAE: Learnable Lattice Vector-Quantization For Efficient Representations