Hi, thanks for this really exciting tool.

I’m trying out some trajectory analysis where I would like to map predicted trajectories from the latent space back to the original expression space; in essence I would like to directly pass the points of these trajectories (which are not part of the anndata object) through the model’s decoder.

I see that the VAE class has a function ‘generative()’ which looks like what I am after, but I am unfamiliar with the architecture of the model and I am unsure how best to access this function. What would be the best way for a user to directly pass latent space points through the model’s decoder?

Any help would be much appreciated, thanks!

Rory