I’m trying to use the multiple Wav2Vec versions with the option of switching seamlessly between one and another.
The “WAVLM_BASE” and “WAVLM_LARGE” version are supposed to be similar but don’t yield the same class.
The WavLM_Base give a Wav2Vec2Model
whereas the WavLM_Large give a _Wav2Vec2Model
.
bundle_base = getattr(torchaudio.pipelines, "WAVLM_BASE")
base_model = bundle_base.get_model()
bundle_large = getattr(torchaudio.pipelines, "WAVLM_LARGE")
large_model = bundle_large.get_model()
assert type(base_model) == type(large_model)
Is this an expected result ?