How to use the pyroomacoustics.datasets.base.Meta function in pyroomacoustics

self.rng.shuffle(rand_idx)
                files = [files[i] for i in rand_idx[:n_files]]
            
            self.size_by_samples[word] = len(files)

            # add each file to the corpus
            for filename in files:

                file_loc = os.path.join(self.basedir, word, os.path.basename(filename))

                # could also add score of original model for each word?
                if speech:
                    meta = Meta(word=word, speech=speech, file_loc=file_loc)
                else:
                    noise_type = os.path.basename(filename).split(".")[0]
                    meta = Meta(word="NA", noise_type=noise_type, speech=speech, file_loc=file_loc)

                if meta.match(**kwargs):
                    self.add_sample(GoogleSample(filename, **meta.as_dict()))

if subset and speech:
                rand_idx = np.arange(len(files)) 
                n_files = min(subset, len(files))
                self.rng.shuffle(rand_idx)
                files = [files[i] for i in rand_idx[:n_files]]
            
            self.size_by_samples[word] = len(files)

            # add each file to the corpus
            for filename in files:

                file_loc = os.path.join(self.basedir, word, os.path.basename(filename))

                # could also add score of original model for each word?
                if speech:
                    meta = Meta(word=word, speech=speech, file_loc=file_loc)
                else:
                    noise_type = os.path.basename(filename).split(".")[0]
                    meta = Meta(word="NA", noise_type=noise_type, speech=speech, file_loc=file_loc)

                if meta.match(**kwargs):
                    self.add_sample(GoogleSample(filename, **meta.as_dict()))

def __init__(self, data, **kwargs):
        ''' Dummy init method '''
        self.data = data
        self.meta = Meta(**kwargs)

def build_corpus(self, **kwargs):
        '''
        Build the corpus with some filters (sex, lang, accent, sentence_tag, sentence)
        '''

        # Check all the sentences
        for tag, info in cmu_arctic_sentences.items():

            # And all speakers for each sentence
            for speaker, path in info['paths'].items():

                # This is the metadata for this sample
                meta = Meta(speaker=speaker, tag=tag, text=info['text'], **cmu_arctic_speakers[speaker])

                # it there is a match, add it
                # The reason we do the match before creating the Sentence object is that
                # we don't want to read the file from disk if there is no match
                if meta.match(**kwargs):
                    self.add_sample(CMUArcticSentence(path, **meta.as_dict()))

self.recordings = self.samples  # this is for convenience
        eval_dir = os.path.join(self.basedir, 'eval')
        dev_dir = os.path.join(self.basedir, 'dev')

        if not os.path.exists(eval_dir) or not os.path.exists(dev_dir):
            warnings.warn('The ''eval'' and/or ''dev'' folders are missing. Please check the structure of the dataset directory.')


        for path, dirs, files in os.walk(self.basedir):

            m = RE_PATH.search(path)

            if m:
                dev = False if m.group(1) == 'eval' else True
                task, rec, array = int(m.group(2)), int(m.group(3)), m.group(4)
                meta = Meta(task=task, rec=rec, array=array, dev=dev)
                if array in locata_arrays and meta.match(**kwargs):
                    if verbose:
                        print(path)
                    self.samples.append(
                            LocataRecording(path, task=task, rec=rec, array=array, dev=dev)
                            )

        if len(self) == 0:
            warnings.warn('Nothing was imported. The dataset can be downloaded at ' + url + '.')