Hi, I wonder if neural compressor supports visual language model that accept visual image and text as inputs?