3a181c8d229d3f45d6457cd329d2336b07b2330b,autokeras/pretrained/voice_generator/voice_generator.py,VoiceGenerator,__init__,#VoiceGenerator#Any#Any#,244

Before Change

        super(VoiceGenerator, self).__init__()
        self.model_path = model_path if model_path is not None else temp_path_generator()
        ensure_dir(self.model_path)
        self.checkpoint_path = os.path.join(self.model_path, Constant.PRE_TRAIN_VOICE_GENERATOR_MODEL_NAME)
        self.sample_rate = 0
        self.hop_length = 0
        self.overwrite = overwrite
        self.device = get_device()

After Change

// Default hyperparameters:

class Hparams:
    name = "deepvoice3"

    // Text:
    // [en  jp]
    frontend = "en"

    // Replace words to its pronunciation with fixed probability.
    // e.g.  "hello" to "HH AH0 L OW1"
    // [en  jp]
    // en: Word -> pronunciation using CMUDict
    // jp: Word -> pronounciation usnig MeCab
    // [0 ~ 1.0]: 0 means no replacement happens.
    replace_pronunciation_prob = 0.5

    // Convenient model builder
    // Definitions can be found at deepvoice3_pytorch/builder.py
    // deepvoice3: DeepVoice3 https://arxiv.org/abs/1710.07654
    builder = "deepvoice3"

Instances