Class SequenceSummary

bert/modeling_utils.py:1069–1143

r""" Compute a single vector summary of a sequence hidden states according to various possibilities: Args of the config class: summary_type: - 'last' => [default] take the last token hidden state (like XLNet) - 'first' => take the first token hidde

class SequenceSummary(nn.Module):
    r""" Compute a single vector summary of a sequence hidden states according to various possibilities:
        Args of the config class:
            summary_type:
                - 'last' => [default] take the last token hidden state (like XLNet)
                - 'first' => take the first token hidden state (like Bert)
                - 'mean' => take the mean of all tokens hidden states
                - 'cls_index' => supply a Tensor of classification token position (GPT/GPT-2)
                - 'attn' => Not implemented now, use multi-head attention
            summary_use_proj: Add a projection after the vector extraction
            summary_proj_to_labels: If True, the projection outputs to config.num_labels classes (otherwise to hidden_size). Default: False.
            summary_activation: 'tanh' or another string => add an activation to the output, Other => no activation. Default
            summary_first_dropout: Add a dropout before the projection and activation
            summary_last_dropout: Add a dropout after the projection and activation
    """

    def __init__(self, config: PretrainedConfig):
        super().__init__()

        self.summary_type = getattr(config, "summary_type", "last")
        if self.summary_type == "attn":
            # We should use a standard multi-head attention module with absolute positional embedding for that.
            # Cf. https://github.com/zihangdai/xlnet/blob/master/modeling.py#L253-L276
            # We can probably just use the multi-head attention module of PyTorch >=1.1.0
            raise NotImplementedError

        self.summary = Identity()
        if hasattr(config, "summary_use_proj") and config.summary_use_proj:
            if hasattr(config, "summary_proj_to_labels") and config.summary_proj_to_labels and config.num_labels > 0:
                num_classes = config.num_labels
            else:
                num_classes = config.hidden_size
            self.summary = nn.Linear(config.hidden_size, num_classes)

        activation_string = getattr(config, "summary_activation", None)
        self.activation: Callable = (get_activation(activation_string) if activation_string else Identity())

        self.first_dropout = Identity()
        if hasattr(config, "summary_first_dropout") and config.summary_first_dropout > 0:
            self.first_dropout = nn.Dropout(config.summary_first_dropout)

        self.last_dropout = Identity()
        if hasattr(config, "summary_last_dropout") and config.summary_last_dropout > 0:
            self.last_dropout = nn.Dropout(config.summary_last_dropout)

    def forward(self, hidden_states, cls_index=None):
        """ hidden_states: float Tensor in shape [bsz, ..., seq_len, hidden_size], the hidden-states of the last layer.
            cls_index: [optional] position of the classification token if summary_type == 'cls_index',
                shape (bsz,) or more generally (bsz, ...) where ... are optional leading dimensions of hidden_states.
                if summary_type == 'cls_index' and cls_index is None:
                    we take the last token of the sequence as classification token
        """
        if self.summary_type == "last":
            output = hidden_states[:, -1]
        elif self.summary_type == "first":
            output = hidden_states[:, 0]
        elif self.summary_type == "mean":
            output = hidden_states.mean(dim=1)