Parallel training of VAE and AAE?

[microbiomix]

Hello,

Would it make sense to allow VAE and AAE to be trained in parallel on request? They are now running one after the other but our server with two GPUs could handle them in parallel. Any thoughts?

Best,
Mani

[simonrasmu]

Hi Mani, very good point - I second that they could in principle run in parallel?

Best
Simon

[microbiomix]

The other option is to run vae model and aae models separately but in parallel, e.g. via Snakemake, and derep downstream. If the behavior using aae-vae is not expected to differ from running them separate (except for non-reproducibility due to random seed etc), perhaps that keeps it simple for avamb code itself? If yes, then we can close this issue or reduce priority to -Infinity.