Browse Source
update
Signed-off-by: Filip Haltmayer <filip.haltmayer@zilliz.com>
main
1 changed files with
2 additions and
2 deletions
-
torchaudio_audio_embedding.py
|
|
|
@ -17,8 +17,8 @@ class TorchaudioAudioEmbedding(Operator): |
|
|
|
self._bundle = getattr(torchaudio.pipelines, name) |
|
|
|
self._model = self._bundle.get_model() |
|
|
|
|
|
|
|
def __call__(self, image_file: 'str') -> NamedTuple('Outputs', [('embedding', numpy.ndarray)]): |
|
|
|
waveform, sample_rate = torchaudio.load(image_file) |
|
|
|
def __call__(self, audio_path: 'str') -> NamedTuple('Outputs', [('embedding', numpy.ndarray)]): |
|
|
|
waveform, sample_rate = torchaudio.load(audio_path) |
|
|
|
waveform = torchaudio.functional.resample(waveform, sample_rate, self._bundle.sample_rate) |
|
|
|
embedding, _ = self._model.extract_features(waveform) |
|
|
|
embedding = embedding[0].detach().numpy() |
|
|
|
|
