Transformers js does not have method post_process_speaker_diarization.

#1
by gowtham2 - opened

Using model.onnx able to get logits but, could do speaker diarization using transformers .js. I am using version 2.17.2 and this is the latest version in github. Any thoughts or suggestions?

ONNX Community org

I am using version 2.17.2 and this is the latest version in github

As stated in the README, this model (and method) is only available in Transformers.js v3 (dev branch), which you can install using:

npm install xenova/transformers.js#v3

Hi, thanks for putting me in right direction. I am able to do use the version3. I am trying to get the number of speakers count in the microphone audio but, I am always getting as 1 even though multiple people are speaking.

ONNX Community org
edited Aug 5

Does the original pyannote segmentation model (in python) work correctly?

@Xenova Yes, it works properly with JavaScript (using bundles). Could you please add support for these features in the CDN in the next release? As I am trying load the model in a web worker, and it would be great to have CDN support for these segmentation models

ONNX Community org
edited Oct 8

You can now access Transformers.js v3 via CDN using:

import { AutoProcessor, AutoModelForAudioFrameClassification, read_audio } from 'https://cdn.jsdelivr.net/npm/@huggingface/[email protected]';

Hope that helps!

Sign up or log in to comment