Duration not preserved?

[maitycyrus]

Why is the duration not preserved during a forward pass through the model?

import dac
from audiotools import AudioSignal
import torch

model_path = dac.utils.download(model_type="24khz")
model = dac.DAC.load(model_path)

x = torch.rand(1, 1, 320*10)
y = model(x)["audio"]

print("duration x = ", x.shape[-1])
print("duration y = ", y.shape[-1])

Output:

duration x =  3200
duration y =  3192