How to use the torchaudio.transforms.Resample function in torchaudio

def test_resample_size(self):
        input_path = os.path.join(self.test_dirpath, 'assets', 'sinewave.wav')
        waveform, sample_rate = torchaudio.load(input_path)

        upsample_rate = sample_rate * 2
        downsample_rate = sample_rate // 2
        invalid_resample = torchaudio.transforms.Resample(sample_rate, upsample_rate, resampling_method='foo')

        self.assertRaises(ValueError, invalid_resample, waveform)

        upsample_resample = torchaudio.transforms.Resample(
            sample_rate, upsample_rate, resampling_method='sinc_interpolation')
        up_sampled = upsample_resample(waveform)

        # we expect the upsampled signal to have twice as many samples
        self.assertTrue(up_sampled.size(-1) == waveform.size(-1) * 2)

        downsample_resample = torchaudio.transforms.Resample(
            sample_rate, downsample_rate, resampling_method='sinc_interpolation')
        down_sampled = downsample_resample(waveform)

        # we expect the downsampled signal to have half as many samples
        self.assertTrue(down_sampled.size(-1) == waveform.size(-1) // 2)

def test_resample_size(self):
        input_path = os.path.join(self.test_dirpath, 'assets', 'sinewave.wav')
        waveform, sample_rate = torchaudio.load(input_path)

        upsample_rate = sample_rate * 2
        downsample_rate = sample_rate // 2
        invalid_resample = torchaudio.transforms.Resample(sample_rate, upsample_rate, resampling_method='foo')

        self.assertRaises(ValueError, invalid_resample, waveform)

        upsample_resample = torchaudio.transforms.Resample(
            sample_rate, upsample_rate, resampling_method='sinc_interpolation')
        up_sampled = upsample_resample(waveform)

        # we expect the upsampled signal to have twice as many samples
        self.assertTrue(up_sampled.size(-1) == waveform.size(-1) * 2)

        downsample_resample = torchaudio.transforms.Resample(
            sample_rate, downsample_rate, resampling_method='sinc_interpolation')
        down_sampled = downsample_resample(waveform)

        # we expect the downsampled signal to have half as many samples
        self.assertTrue(down_sampled.size(-1) == waveform.size(-1) // 2)

def test_vctk_transform_pipeline(self):
        test_filepath_vctk = os.path.join(self.test_dirpath, "assets/VCTK-Corpus/wav48/p224/", "p224_002.wav")
        wf_vctk, sr_vctk = torchaudio.load(test_filepath_vctk)

        # rate
        sample = T.Resample(sr_vctk, 16000, resampling_method='sinc_interpolation')
        wf_vctk = sample(wf_vctk)
        # dither
        wf_vctk = F.dither(wf_vctk, noise_shaping=True)

        E = torchaudio.sox_effects.SoxEffectsChain()
        E.set_input_file(test_filepath_vctk)
        E.append_effect_to_chain("gain", ["-h"])
        E.append_effect_to_chain("channels", [1])
        E.append_effect_to_chain("rate", [16000])
        E.append_effect_to_chain("gain", ["-rh"])
        E.append_effect_to_chain("dither", ["-s"])
        wf_vctk_sox = E.sox_build_flow_effects()[0]

        self.assertTrue(torch.allclose(wf_vctk, wf_vctk_sox, rtol=1e-03, atol=1e-03))

def __init__(self, transforms=None, shiftRandom=True, validationAudioPath=None):
        self.transforms = transforms
        self.preview = validationAudioPath is not None
        self.shiftRandom = shiftRandom and not self.preview
        self.count = None

        animFPS = 29.97  # samSoar recorded with an ipad

        if self.preview:
            inputSpeechPath = validationAudioPath
        else:
            inputSpeechPath = os.path.join(ROOT_PATH, 'data', 'samSoar', 'samSoar.wav')
        self.waveform, self.sampleRate = torchaudio.load(inputSpeechPath)
        if self.sampleRate != 16000:
            self.waveform = torchaudio.transforms.Resample(self.sampleRate, 16000)(self.waveform)
            self.sampleRate = 16000

        self.count = int(animFPS * (self.waveform.size()[1] / self.sampleRate))

        self.LPC = lpc.LPCCoefficients(
            self.sampleRate,
            .032,
            .5,
            order=31  # 32 - 1
        )

        if os.path.exists(INPUT_VALUES_PRECALC_PATH):
            self.inputValues = torch.load(INPUT_VALUES_PRECALC_PATH)
        else:
            print('pre-calculating input values...')
            self.inputValues = torch.Tensor([])

if not self.training:
                unmix_target = model.load_model(
                    target=target,
                    model_name=self.model_name,
                    device=self.device
                )
                unmix_target.freeze()
            else:
                unmix_target = self.target_models[j]
            model_rate = unmix_target.sample_rate.item()

            if rate != model_rate:
                # we have to resample to model samplerate if needed
                # this makes sure we resample input only once
                model_rate = unmix_target.sample_rate.item()
                resampler = torchaudio.transforms.Resample(
                    orig_freq=rate,
                    new_freq=model_rate).to(self.device)
                # Until torchaudio has not merged the PR
                # https://github.com/pytorch/audio/pull/277 , There is a need
                # to do the resampling on cpu
                audio = torch.cat(
                    [resampler(audio[sample])[None, ...].cpu()
                     for sample in range(nb_samples)],
                    dim=0).to(self.device)
                rate = model_rate

            # apply current model to get the source spectrogram
            Vj = unmix_target(audio)

            if self.softmask:
                # only exponentiate the model if we use softmask

class MFCC:
    forward = torchaudio.transforms.MFCC().forward


class MuLawEncoding:
    forward = torchaudio.transforms.MuLawEncoding().forward


class MuLawDecoding:
    forward = torchaudio.transforms.MuLawDecoding().forward


class Resample:
    # Resample isn't a script_method
    forward = torchaudio.transforms.Resample.forward

def __init__(self, orig_freq=16000, new_freq=16000, resampling_method='sinc_interpolation'):
        super(Resample, self).__init__()
        self.orig_freq = orig_freq
        self.new_freq = new_freq
        self.resampling_method = resampling_method