Thank you for your work!
Hey, your work with the base models is really titanic! I really appreciate that you have included Spanish among the languages!
To train your model, do you use Linux or Windows? Local or Cloud VM machine? How much RAM do you have?
I myself tried these last days to make a mini “base” model (although it's really a fine-tuning of the RVC default models) in Spanish with 30 hours of studio quality audios from all the Latin American region, so you can find a variety of accents.
Only that even having a 3090 and 64 GB of ram, even before starting to train the shared memory of the system was filled causing an error, I could not even know if the VRAM was also filled as it marked 1 GB / 24 GB, (Anyway I think I read a CUDA out of memory error on console. No matter what batch I put in.) and the shared memory was increasing more and more until it crashed.
From what I read, the shared memory uses half of your RAM and cannot be modified, only by adding more RAM. Thanks, Windows… ಠ_ಠ
Cheers!