Process any prompt (limited to 512 bytes) with the 100M entropy patcher model and visualize the token entropies plot below.NOTE: this implementation differs slightly by excluding local attention so we limit the characters limit to 512 to avoid any deviation.