About remesher of MeshGraphNets ?

[chufall]

Hi, I'm reading the paper "LEARNING MESH-BASED SIMULATION WITH GRAPH NETWORKS".
In the section 3.2 "ADAPTIVE REMESHING", the paper metioned "At test time, for each time step we predict both the next simulation state and the sizing field, and use a generic, domain- independent remesher R to compute the adapted next-step mesh".
There're two questions:

1. Does the remeshing operation will changed the number of nodes? since one of the remeshing is spliting which will create a new node?
2. Will it perform the remeshing operation when traing the model?

Thanks a lot!
qc