search

nttcslab / m2d

Masked Modeling Duo: Towards a Universal Audio Pre-training Framework
https://ieeexplore.ieee.org/document/10502167
Other
54 stars 1 forks source link

Questions about model build-up #1

Closed SaoYear closed 1 year ago

SaoYear commented 1 year ago

Hi there,

Thanks for the open-source codes and this fantistic work!

I am reproducing this work now, but I met a problem in the model build-up. Specifically, in the runtime_audio.py, line 84.

model = models_mae.__dict__[args.model](img_size=args.input_size, patch_size=args.patch_size, decoder_depth=args.decoder_depth)

I load the models_mae.py according to the intro, but the args.model = m2d_vit_base after the parent folder name parsing, which is not provided in the vanilla models_mae.py file.

So I would like to know how to import the model in a correct way...

Thanks in advance!

SaoYear commented 1 year ago

I did not finish the intro... I see the problem.

daisukelab commented 1 year ago

@SaoYear Hello, thank you for your interest. I understand that your issue might have been resolved by following the steps written in the README.md. Please feel free to ask if you find another problem in the future. Thanks again!