How can I use MULTI-PHASE GAMMATONE FILTERBANK?

[seacj]

Here's my code to generate a gammatone spectrogram. But it seems weird. So I want know the correct way to use it.

import soundfile as sf
from asteroid.filterbanks import STFTFB
from asteroid.filterbanks.enc_dec import Filterbank, Encoder, Decoder

audio_path = 'xx.wav'
y, sr = sf.read(audio_path)

dft_filters = STFTFB(n_filters=512, kernel_size=256, stride=128, sample_rate=16000, window = torch.hamming_window(256).numpy())
stft = Encoder(dft_filters)
data = stft(torch.Tensor([y]))
print(data.shape)
plot_spectrogram(data[0][:256].detach().numpy())

dft_filters = MultiphaseGammatoneFB()
stft = Encoder(dft_filters)
data = stft(torch.Tensor([y]))
print(data.shape)
plot_spectrogram(data[0].detach().numpy())

[mpariente]

I'm sure @dditter can tell if this is correct or not ;-)