Can you please explain each input in details.

#1
by tahercoolguy - opened

Thanks for opensourcing this model.

Your output from this model for closest is awesome . But other thing input things are really not understandable . It Will be fantastic if you could explain each input value and behaviour it will impart in generation.

If you don't have enough time for it . So just give vague explanation of each input . I will do research and send you all detailed instruction set so that you can published on your main page .

Thank you

Owner

Hi there, the inputs to my model are the same as those to MusicGen from the audiocraft repo on GitHub. some docstrings on generation params are here. The functions I use in the code from the model card are in that file too, so you can find a little more info on those as well.

The model is a language model essentially, so you'll see similar args to what you might see for llama, etc (top_p, top_k, ...).

I'll write up a more detailed post on this and share it soon. will ping ya here when I do and will add more info to model card. If you have any specific inputs you're unsure about, or other specific questions please lmk I'll do my best to respond quickly

Thank for replying will see audiocraft and generation params from llama .

Owner

Did write a blogpost but its more focused on motivation/training details/data/the app I put together, not inference params.

Either way I think I'll close this for now.

nateraw changed discussion status to closed

Sign up or log in to comment