is it typo in UnitExtractor? self.collate = Collater(pad_value=2, pad_to_multiple=2)

[seastar105]

seamless_communication/src/seamless_communication/models/unit_extractor/unit_extractor.py

Lines 37 to 59 in 91918a6

	class UnitExtractor(nn.Module):
	"""Unit Extractor which converts raw audio into units."""
	def __init__(
	self,
	model_name_or_card: Union[str, AssetCard],
	kmeans_uri: str,
	device: Device,
	dtype: DataType = torch.float32,
	):
	super().__init__()
	wav2vec2_model = load_wav2vec2_model(
	model_name_or_card, device=device, dtype=dtype
	)
	wav2vec2_model.eval()
	assert isinstance(wav2vec2_model, Wav2Vec2Model)
	self.model = Wav2Vec2LayerOutputModel(wav2vec2_model)
	self.decode_audio = AudioDecoder(dtype=torch.float32, device=device)
	self.collate = Collater(pad_value=2, pad_to_multiple=2)
	self.kmeans_model = KmeansModel(kmeans_uri, device, dtype)
	self.device = device
	self.dtype = dtype

UnitExtractor collates batch using Collater(pad_value=2, pad_to_multiple=2). it seems unusual to use pad_value as 2. is it correct? or typo?

[kauterry]

You are right, it is incorrect, I fix it in #251