Using Wav2Vec2 model

[evelynlee999]

Thanks for your code,it helped me a lot,but I tried to use Wav2Vec2 model and got error as following:
RuntimeError: Expected 2D (unbatched) or 3D (batched) input to conv1d, but got input of size: [32, 1, 301, 40]
It showed the error at output.
`def wav2vec2_embed_raw_audio(input_tensor: t.Tensor, model: Wav2Vec2Model) -> t.Tensor:

output: Wav2Vec2BaseModelOutput = model(input_tensor)

features = output.last_hidden_state
features = features.transpose(1, 2)

return features`

Should I transform t.Tensor into the path of voxceleb or it should be t.tensor.

And I have another problem,download_pretrained_models.sh downloaded some .pt files,but it seemed to be useless,and I don't know where to use them.