def __init__(self,
audiopaths_and_text,
hparams,
speaker_ids=None,
mode='train'):
self.hparams = hparams
tmp = mode.split('-')
if tmp[0] == 'train':
self.audiopaths_and_text = load_filepaths_and_text_train(
audiopaths_and_text, split='\t')
if len(tmp) == 2:
self.mode = tmp[1]
else:
self.mode = True
else:
if isinstance(audiopaths_and_text,
(str, Path)) and os.path.isfile(audiopaths_and_text):
self.audiopaths_and_text = load_filepaths_and_text(
audiopaths_and_text, split='\t')
else:
self.audiopaths_and_text = ['audiopath', 'text', 'speaker']
self.mode = False
self.text_cleaners = hparams.text_cleaners
self.max_wav_value = hparams.max_wav_value
self.sampling_rate = hparams.sampling_rate
self.stft = layers.TacotronSTFT(hparams.filter_length,
hparams.hop_length, hparams.win_length,
hparams.n_mel_channels,
hparams.sampling_rate,
hparams.mel_fmin, hparams.mel_fmax)
self.sampling_rate = hparams.sampling_rate
self.filter_length = hparams.filter_length
self.hop_length = hparams.hop_length
self.f0_min = hparams.f0_min
self.f0_max = hparams.f0_max
self.harm_thresh = hparams.harm_thresh
self.p_arpabet = hparams.p_arpabet
self.max_decoder_steps = hparams.max_decoder_steps
self.f0_dim = hparams.prenet_f0_dim # f0的维度设置
self.encoder_model_fpath = hparams.encoder_model_fpath
self.cmudict = None
if hparams.cmudict_path is not None:
self.cmudict = cmudict.CMUDict(hparams.cmudict_path)
self.speaker_ids = speaker_ids
if self.speaker_ids is None:
self.speaker_ids = self.create_speaker_lookup_table(
self.audiopaths_and_text)
# random.seed(1234)
# random.shuffle(self.audiopaths_and_text)
self.ids = set(range(len(self.audiopaths_and_text)))
def __init__(self,
audiopaths_and_text,
hparams,
speaker_ids=None,
mode='train'):
self.hparams = hparams
tmp = mode.split('-')
if tmp[0] == 'train':
self.audiopaths_and_text = load_filepaths_and_text_train(
audiopaths_and_text, split='\t')
if len(tmp) == 2:
self.mode = tmp[1]
else:
self.mode = True
else:
self.audiopaths_and_text = load_filepaths_and_text(
audiopaths_and_text, split='\t')
self.mode = False
self.text_cleaners = hparams.text_cleaners
self.max_wav_value = hparams.max_wav_value
self.sampling_rate = hparams.sampling_rate
self.stft = layers.TacotronSTFT(hparams.filter_length,
hparams.hop_length, hparams.win_length,
hparams.n_mel_channels,
hparams.sampling_rate,
hparams.mel_fmin, hparams.mel_fmax)
self.sampling_rate = hparams.sampling_rate
self.filter_length = hparams.filter_length
self.hop_length = hparams.hop_length
self.f0_min = hparams.f0_min
self.f0_max = hparams.f0_max
self.harm_thresh = hparams.harm_thresh
self.p_arpabet = hparams.p_arpabet
self.f0_dim = hparams.prenet_f0_dim # f0的维度设置
self.cmudict = None
if hparams.cmudict_path is not None:
self.cmudict = cmudict.CMUDict(hparams.cmudict_path)
self.speaker_ids = speaker_ids
if self.speaker_ids is None:
self.speaker_ids = self.create_speaker_lookup_table(
self.audiopaths_and_text)