facebookresearch / audiocraft

Audiocraft is a library for audio processing and generation with deep learning. It features the state-of-the-art EnCodec audio compressor / tokenizer, along with MusicGen, a simple and controllable music generation LM with textual and melodic conditioning.
MIT License
20.92k stars 2.14k forks source link

[research question] does the model predict the last few codes in delayed pattern #95

Open jasonppy opened 1 year ago

jasonppy commented 1 year ago

Hi,

Thanks for the great work and open sourcing everything!

On the delayed pattern

image

it seems that there should be a few model tokens at the end to be predicted (I drew some of them out on the left). Are those predicted? or ignoring those doesn't affect result (which is reasonable considering the first codebook is the most important and only a very small amount of codes from higher codebooks are dropped)

Also, Looking forward to the training code!

Thanks