--latent_size parameter not present in export.py

[mo-seph]

The tutorial (https://forum.ircam.fr/article/detail/training-rave-models-on-custom-data/) says that --latent_size passed to rave encode sets the number of latent dimensions when exporting for nn~ etc. However, it doesn't work (Flags parsing error: Unknown command line flag 'latent_size') and the code in export.py does not seem to have it as a parsable argument. fidelity is there, but I'd like to be sure how many dimensions the model has in order to be able to patch consistently

[audiophil-dev]

I can confirm this. As a workaround I tested different fidelity parameter vaues until I got the right number of latents.