Add ReDimNet embedding model

[rose-jinyang]

Hello
How are you?
Thanks for contributing to this project.
I am trying to add ReDimNet model (https://github.com/IDRnD/ReDimNet) as embedding model of piple-line.
But the ReDimNet model does not require weights parameter when getting embeddings from audio signal data.
How can I add this model?
Thanks

[juanmc2005]

Hi @rose-jinyang ! I love the idea of adding compatibility with new models! I would gladly merge a PR for this if you're up for it.
I suggest you take a look at how pyannote.audio implements weights for models such as WeSpeaker or SpeechBrain embeddings.
If I remember correctly, you can use the weights as a mask over the audio (one mask per speaker) and extract the embedding for each speaker from the masked audio. This will allow you to output one embedding per speaker by taking audio and weights as input.

Feel free to open a draft PR so I can guide you through it!