Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Regression Task #110

Open
BaMarcy opened this issue Aug 18, 2023 · 1 comment
Open

Regression Task #110

BaMarcy opened this issue Aug 18, 2023 · 1 comment
Labels
question Further information is requested

Comments

@BaMarcy
Copy link

BaMarcy commented Aug 18, 2023

Hello,

I'm interested in knowing whether this model can be utilized for regression tasks. From my analysis of the architecture, it appears that the model incorporates nn.Linear() towards the final layer, which leads me to believe that regression tasks might be supported. However, I'd greatly appreciate some clarification to ensure my understanding is accurate.

Thanks!

@YuanGongND YuanGongND added the question Further information is requested label Aug 18, 2023
@YuanGongND
Copy link
Owner

YuanGongND commented Aug 18, 2023

hi there,

Yes, AST can do regression, you don't even need to change the model architecture, just set label_dim=1 when you instantiate the model, so the output would be a single value.

def __init__(self, label_dim=527, fstride=10, tstride=10, input_fdim=128, input_tdim=1024, imagenet_pretrain=True, audioset_pretrain=False, model_size='base384', verbose=True):

You however need to change the training pipeline to use a different loss such as MSE, etc.

I cannot guarantee anything beyond "the provided code can reproduce the results shown in the paper". But in my opinion, AST should at least get similar regression results as other models.

-Yuan

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
question Further information is requested
Projects
None yet
Development

No branches or pull requests

2 participants