Method __init__

modules/fastspeech/tts_modules.py:311–322

(self, embed_tokens, hidden_size=None, num_layers=None, kernel_size=None, num_heads=2)

class FastspeechEncoder(FFTBlocks):
    def __init__(self, embed_tokens, hidden_size=None, num_layers=None, kernel_size=None, num_heads=2):
        hidden_size = hparams['hidden_size'] if hidden_size is None else hidden_size
        kernel_size = hparams['enc_ffn_kernel_size'] if kernel_size is None else kernel_size
        num_layers = hparams['dec_layers'] if num_layers is None else num_layers
        super().__init__(hidden_size, num_layers, kernel_size, num_heads=num_heads,
                         use_pos_embed=False)  # use_pos_embed_alpha for compatibility
        self.embed_tokens = embed_tokens
        self.embed_scale = math.sqrt(hidden_size)
        self.padding_idx = 0
        self.embed_positions = SinusoidalPositionalEmbedding(
            hidden_size, self.padding_idx, init_size=DEFAULT_MAX_TARGET_POSITIONS,
        )

    def forward(self, txt_tokens):
        """