Extract embeddings for speaker identification?

[cbowdon]

Hi! This is a great library, thanks for open sourcing it.

Is it possible to extract embeddings from this model that can then be clustered for speaker identification? E.g. could I take the output of the encoder here before the combined embedding is created?

SpeechLLM/huggingface/hf_repo/model.py

Line 68 in f44d361

speech_embeds = self.audio_encoder(speech)

I'm new to speech processing so please forgive me if that's daft. Thanks!