Class AttnAddedKVProcessor

models_diffusers/attention_processor.py:873–934

r""" Processor for performing attention-related computations with extra learnable key and value matrices for the text encoder.

class AttnAddedKVProcessor:
    r"""
    Processor for performing attention-related computations with extra learnable key and value matrices for the text
    encoder.
    """

    def __call__(
        self,
        attn: Attention,
        hidden_states: torch.FloatTensor,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        scale: float = 1.0,
    ) -> torch.Tensor:
        residual = hidden_states

        args = () if USE_PEFT_BACKEND else (scale,)

        hidden_states = hidden_states.view(hidden_states.shape[0], hidden_states.shape[1], -1).transpose(1, 2)
        batch_size, sequence_length, _ = hidden_states.shape

        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)

        if encoder_hidden_states is None:
            encoder_hidden_states = hidden_states
        elif attn.norm_cross:
            encoder_hidden_states = attn.norm_encoder_hidden_states(encoder_hidden_states)

        hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)

        query = attn.to_q(hidden_states, *args)
        query = attn.head_to_batch_dim(query)

        encoder_hidden_states_key_proj = attn.add_k_proj(encoder_hidden_states, *args)
        encoder_hidden_states_value_proj = attn.add_v_proj(encoder_hidden_states, *args)
        encoder_hidden_states_key_proj = attn.head_to_batch_dim(encoder_hidden_states_key_proj)
        encoder_hidden_states_value_proj = attn.head_to_batch_dim(encoder_hidden_states_value_proj)

        if not attn.only_cross_attention:
            key = attn.to_k(hidden_states, *args)
            value = attn.to_v(hidden_states, *args)
            key = attn.head_to_batch_dim(key)
            value = attn.head_to_batch_dim(value)
            key = torch.cat([encoder_hidden_states_key_proj, key], dim=1)
            value = torch.cat([encoder_hidden_states_value_proj, value], dim=1)
        else:
            key = encoder_hidden_states_key_proj
            value = encoder_hidden_states_value_proj

        attention_probs = attn.get_attention_scores(query, key, attention_mask)
        hidden_states = torch.bmm(attention_probs, value)
        hidden_states = attn.batch_to_head_dim(hidden_states)

        # linear proj
        hidden_states = attn.to_out[0](hidden_states, *args)
        # dropout
        hidden_states = attn.to_out[1](hidden_states)