Method __init__

inspiremusic/transformer/encoder.py:344–387

Construct TransformerEncoder See Encoder for the meaning of each parameter.

(
        self,
        input_size: int,
        output_size: int = 256,
        attention_heads: int = 4,
        linear_units: int = 2048,
        num_blocks: int = 6,
        dropout_rate: float = 0.1,
        positional_dropout_rate: float = 0.1,
        attention_dropout_rate: float = 0.0,
        input_layer: str = "conv2d",
        pos_enc_layer_type: str = "abs_pos",
        normalize_before: bool = True,
        static_chunk_size: int = 0,
        use_dynamic_chunk: bool = False,
        global_cmvn: torch.nn.Module = None,
        use_dynamic_left_chunk: bool = False,
        key_bias: bool = True,
        selfattention_layer_type: str = "selfattn",
        activation_type: str = "relu",
        gradient_checkpointing: bool = False,
    )

    """Transformer encoder module."""

    def __init__(
        self,
        input_size: int,
        output_size: int = 256,
        attention_heads: int = 4,
        linear_units: int = 2048,
        num_blocks: int = 6,
        dropout_rate: float = 0.1,
        positional_dropout_rate: float = 0.1,
        attention_dropout_rate: float = 0.0,
        input_layer: str = "conv2d",
        pos_enc_layer_type: str = "abs_pos",
        normalize_before: bool = True,
        static_chunk_size: int = 0,
        use_dynamic_chunk: bool = False,
        global_cmvn: torch.nn.Module = None,
        use_dynamic_left_chunk: bool = False,
        key_bias: bool = True,
        selfattention_layer_type: str = "selfattn",
        activation_type: str = "relu",
        gradient_checkpointing: bool = False,
    ):
        """ Construct TransformerEncoder

        See Encoder for the meaning of each parameter.
        """
        super().__init__(input_size, output_size, attention_heads,
                         linear_units, num_blocks, dropout_rate,
                         positional_dropout_rate, attention_dropout_rate,
                         input_layer, pos_enc_layer_type, normalize_before,
                         static_chunk_size, use_dynamic_chunk, global_cmvn,
                         use_dynamic_left_chunk, gradient_checkpointing)
        activation = INSPIREMUSIC_ACTIVATION_CLASSES[activation_type]()
        self.encoders = torch.nn.ModuleList([
            TransformerEncoderLayer(
                output_size,
                INSPIREMUSIC_ATTENTION_CLASSES[selfattention_layer_type](attention_heads,
                                                                      output_size,
                                                                      attention_dropout_rate,
                                                                      key_bias),
                PositionwiseFeedForward(output_size, linear_units,
                                        dropout_rate, activation),
                dropout_rate, normalize_before) for _ in range(num_blocks)
        ])


class ConformerEncoder(BaseEncoder):