Created by: qo4on

I found that LJSpeech class reads wav files from the dataset using _get_example function from the Dataset class. Can anybody explain to me how this works, especially what do these three lines do:

validloader = DataCargo(validset, batch_size=1, shuffle=False)
validreader = fluid.io.PyReader(capacity=20, return_list=True)
validreader.decorate_batch_generator(validloader, place)

class Dataset(ljspeech.LJSpeech):
    def __init__(self, config):
        super(Dataset, self).__init__(config.root)
        self.config = config

    def _get_example(self, metadatum):
        fname, _, _ = metadatum
        wav_path = os.path.join(self.root, "wavs", fname + ".wav")

        loaded_sr, audio = read(wav_path)
        assert loaded_sr == self.config.sample_rate
        return audio

class LJSpeech:
    def __init__(self, config, nranks, rank):
        place = fluid.CUDAPlace(rank) if config.use_gpu else fluid.CPUPlace()

        # Whole LJSpeech dataset.
        ds = Dataset(config)
        
        # Split into train and valid dataset.
        indices = list(range(len(ds)))
        
        valid_indices = indices[:config.valid_size]

        # Valid dataset.
        validset = Subset(ds, valid_indices, valid=True)
        # Currently only support batch_size = 1 for valid loader.
        validloader = DataCargo(validset, batch_size=1, shuffle=False)
        validreader = fluid.io.PyReader(capacity=20, return_list=True)
        validreader.decorate_batch_generator(validloader, place)
        self.validloader = validreader