Finetuning the model

[greycellz]

Hello,

This is outstanding work. Congratulations to everyone involved.

I'm trying to figure out a way to fine tune the model so it can create some other language specific songs. Is there some direction on how to train the model on a selected dataset? Can it be done on hugging face?

Thanks!

[WrongProtocol]

They will be publishing a whitepaper which will document the process for fine-tuning.

[a43992899]

They will be publishing a whitepaper which will document the process for fine-tuning.

Actually, the paper will mostly be talking about the pretraining process. But finetuning shares a similar process. I will try to document one of our fine-tuning experiments in the neural plasticity section.

[greycellz]

Thank you! That will be awesome, I'm looking forward to it.

One quick question: I am trying to build a colab notebook that could potentially be used by users of the model to go step by step. I think I'm almost there but getting an issue at the last step. I wonder if anyone could take a quick look and help figure out what the problem with the "model" import in the last step could be. I can see that the model directory is already downloaded from HF but seems Infer is not able to find the module. Any help will be awesome! I'll add the Colab piece and instructions to running with GPUs on Colab if I'm able to get help here. Thanks!

https://drive.google.com/file/d/1YhT6b1FT6fK7gNgPWryZemUExJFhYwFw/view?usp=sharing

[WrongProtocol]

this seems like a path issue.
I haven't run the notebook at this time, but
/content/YuE/inference/xcodec_mini_infer/models seems to be the path you want

i see you're referencing a path to /content/xcodec_mini_infer , /content/inference .... which both seem to be invalid paths
but not /content/YuE/inference/xcodec_mini_infer/models

hope that solves it.