Question about the use of appearance code

[ChikaYan]

Hi and thank you for the brilliant work!

While reading the code base, I found out the HyperNerf.use_nerf_embed is set to false in all of the given configurations. Based on my understanding this flag is used to determine whether additional embedding, such as appearance code, is added as input to the NeRF network. This seems to contradict the paper which suggests a per-frame appearance code is fed into the network along with view directions. May I ask for what types of experiments do you use the appearance code? And does it constantly give better performance? Many thanks.