snakers4
commented
1 week ago https://github.com/snakers4/silero-vad/issues/474
will be fixed here we know about the problem
the choice of the window was a cool feature that allowed me to optimize the detection
turns out some people were passing the whole instead of a chunk in any case with the new version we changed some internals in the model and audio chunk size choice became pointless since the results now hardly depend on it
Hi, I've been using your model to detect vocals in audio tracks. It was working pretty fine and the choice of the window was a cool feature that allowed me to optimize the detection. Now with the newer version I see a lot of the activity is not detected, and I would revert back to the older model. Is there a way to do this? I tried loading the older model from the hub but it seems it's not available anymore.
By loading it in this way: USE_ONNX = False # change this to True if you want to test onnx model if USE_ONNX: !pip install -q onnxruntime
model, utils = torch.hub.load(repo_or_dir='snakers4/silero-vad:v4.0', model='silero_vad', force_reload=True, onnx=USE_ONNX)
(get_speech_timestamps, save_audio, read_audio, VADIterator, collect_chunks) = utils
I get this error: ~/.cache/torch/hub/helloWorld199_silero-vadv4_master/hubconf.py in
2 import torch
3 import json
----> 4 from utils_vad import (init_jit_model,
5 get_speech_timestamps,
6 get_number_ts,
ImportError: cannot import name 'get_number_ts' from 'utils_vad' (/root/.cache/torch/hub/snakers4_silero-vad_master/utils_vad.py)
I'm don't really know how torch hub works, but I would like to know how to resolve this versions conflict, and use the v4 one,
Thank you!