About train Dac model

[RevolGMPHL]

The current compression model is 32khz mono. I'm wondering if you guys can release a model trained using a 44khz DAC to compress, or one trained on an even higher quality codec, to compress audio. It's really hard to train models like nq9 and nq18. Thanks!

[Permafacture]

There is a trained 44khx DAC in this repo.

model = CompressionModel.get_pretrained('dac_44khz')

Looks like this repo does not support training of it but this does: https://github.com/descriptinc/descript-audio-codec

[alba-saco]

I am also hoping for a released model trained using 44khz DAC or 48khz fb encodec