Deep learning expert needed for normalizing flows implementation

Deep learning expert needed for implementing few normalizing flows on the custom data

So the project is to implement this on my data with various flows and transformations in it
https://github.com/didriknielsen/survae_flows
The example on the main page I have applied that transformation(affine coupling bijection) . It is not giving good results. I want to try other transformations as well now as given in the GitHub
other two as well and importantly these different transformations on flow model
https://github.com/didriknielsen/survae_flows/tree/master/survae/transforms
and these as well
https://github.com/didriknielsen/survae_flows/tree/master/experiments/toy
and this. It is all in the main repo
https://github.com/didriknielsen/survae_flows/tree/master/experiments/point_cloud
I need the final output in the csv format and latent representation as well in csv
so the thing Is the data loader should read any csv file with the same matrix form and I should get the output files as mentioned :)

and optimized models as well to get the good results

ABOUT DATASET(ATTACHED):
It is cell by gene matrix
each cell is 1 by 3000 dimension and total cells are 600