Shape error while predicting on custom video

[Zepharchit]

Hello,
Thanks for the amazing work.
I was trying action recognition using MS-G3D repository.
While training I had the following model parameters:-
model_args :
num_class:6
num_points:18
num_person:2
num_gcn_scales:8
num_g3d_scales:8
graph:graph.kinetics.AdjMatrixGraph

But when I am testing the model I have a data shape of N,C,T,V,M as [1,3,128,18,1]
I get an error of rolling mean should be 54 but it is 108 in self.data_bn(x) - msg3d.py(line 156)

Requesting help on how to do prediction when only 1 person is in the data