Add TUNet

[OllieBroadhurst]

🌟 New model addition

The TUNet model for audio superresolution

Model description

TUNet: A Block-online Bandwidth Extension Model based on Transformers and Self-supervised Pretraining is a paper by Viet-Anh Nguyen, Anh H. T. Nguyen, and Andy W. H. Khong, introducing a model for audio superresolution based on transformers.

Audio superresolution allows for the upsampling of audio with minimal loss of quality. This is very useful for ASR tasks that require audio to be resampled during preprocessing, which has a big impact on transcriptions depending on the native sample rate.

Open source status

• [x] the model implementation is available: The official repo
• [x] the model weights are available: ONNX weights
• [x] who are the authors: Viet-Anh Nguyen, Anh H. T. Nguyen, and Andy W. H. Khong

[patil-suraj]

cc @patrickvonplaten @anton-l

[02shanks]

Hi @patrickvonplaten @anton-l @NielsRogge @NielsRogge @amyeroberts @patil-suraj Is this model still available for contribution? I would like to contribute to this model. Can you please guide me! Looking forward from hearing from you all!

[amyeroberts]

Hi @02shanks,

We have recently been trying to push for model on the hub and have as much support as we can there. This is the recommended way of adding new models and it will also be easier to integrate it! Here is a tutorial if that sound good to you!

cc @sanchit-gandhi

[02shanks]

yaa sounds good to me. Will go through tutorial and try to implement it!