Class MLPEncoder

toolkit/models/modules/encoder.py:9–41

The subnetwork that is used in TFN for video and audio in the pre-fusion stage

## 这两个模块都是用在 TFN 中的 (video|audio)
class MLPEncoder(nn.Module):
    '''
    The subnetwork that is used in TFN for video and audio in the pre-fusion stage
    '''

    def __init__(self, in_size, hidden_size, dropout):
        '''
        Args:
            in_size: input dimension
            hidden_size: hidden layer dimension
            dropout: dropout probability
        Output:
            (return value in forward) a tensor of shape (batch_size, hidden_size)
        '''
        super(MLPEncoder, self).__init__()
        # self.norm = nn.BatchNorm1d(in_size)
        self.drop = nn.Dropout(p=dropout)
        self.linear_1 = nn.Linear(in_size, hidden_size)
        self.linear_2 = nn.Linear(hidden_size, hidden_size)
        self.linear_3 = nn.Linear(hidden_size, hidden_size)

    def forward(self, x):
        '''
        Args:
            x: tensor of shape (batch_size, in_size)
        '''
        # normed = self.norm(x)
        dropped = self.drop(x)
        y_1 = F.relu(self.linear_1(dropped))
        y_2 = F.relu(self.linear_2(y_1))
        y_3 = F.relu(self.linear_3(y_2))

        return y_3


# TFN 中的文本编码，额外需要lstm 操作 [感觉是audio|video]