TorchAudio : __init__() got an unexpected keyword argument 'center'

audio
1

Hello,

I was passing some mel arguments into the torchaudio.transforms.MFCC function but it doesn’t seem to recognize the argument “center”. Center is a parameter to the Mel Spectorgram so I don’t understand why it’s not working.

Code:

melkwargs = {
    "center": False,
    "n_fft": n_fft,
    "hop_length": hop_length,
    "power": 2,
    "f_min": f_min,
    "f_max": f_max
}

mfcc_module = torchaudio.transforms.MFCC(sample_rate=sample_rate, n_mfcc=20, melkwargs=melkwargs)
torch_mfcc = mfcc_module(torch.tensor(waveform))

Error:

TypeError: init() got an unexpected keyword argument ‘center’

Any help would be appreciated.

2

You might need to update torchaudio to 0.8.0, as the center argument doesn’t seem to be available in previous versions.