pytext/models/decoders/mlp_decoder_n_tower.py [68:100]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.out_dim = to_dim
        self.export_type = export_type
        log_class_usage

    @staticmethod
    def get_mlp(
        from_dim: int,
        to_dim: int,
        hidden_dims: List[int],
        layer_norm: bool,
        dropout: float,
        activation: Activation,
        export_embedding: bool = False,
    ):
        layers = []
        for i in range(len(hidden_dims)):
            dim = hidden_dims[i]
            layers.append(nn.Linear(from_dim, dim, True))
            # Skip ReLU, LayerNorm, and dropout for the last layer if export_embedding
            if not (export_embedding and i == len(hidden_dims) - 1):
                layers.append(get_activation(activation))
                if layer_norm:
                    layers.append(nn.LayerNorm(dim))
                if dropout > 0:
                    layers.append(nn.Dropout(dropout))
            from_dim = dim

        if to_dim > 0:
            layers.append(nn.Linear(from_dim, to_dim, True))

        return nn.Sequential(*layers)

    def forward(self, *x: List[torch.Tensor]) -> torch.Tensor:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



pytext/models/decoders/mlp_decoder_two_tower.py [120:152]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.out_dim = to_dim
        self.export_type = export_type
        log_class_usage

    @staticmethod
    def get_mlp(
        from_dim: int,
        to_dim: int,
        hidden_dims: List[int],
        layer_norm: bool,
        dropout: float,
        activation: Activation,
        export_embedding: bool = False,
    ):
        layers = []
        for i in range(len(hidden_dims)):
            dim = hidden_dims[i]
            layers.append(nn.Linear(from_dim, dim, True))
            # Skip ReLU, LayerNorm, and dropout for the last layer if export_embedding
            if not (export_embedding and i == len(hidden_dims) - 1):
                layers.append(get_activation(activation))
                if layer_norm:
                    layers.append(nn.LayerNorm(dim))
                if dropout > 0:
                    layers.append(nn.Dropout(dropout))
            from_dim = dim

        if to_dim > 0:
            layers.append(nn.Linear(from_dim, to_dim, True))

        return nn.Sequential(*layers)

    def forward(self, *x: List[torch.Tensor]) -> torch.Tensor:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -