attention_processor.py

from diffusers.models.attention import Attention
from typing import Optional
import torch
import numpy as np


class AttnProcessor:
    """
    Default processor for performing attention-related computations.
    """
    def __init__(self):
        # Initialize an empty list to store attention maps for each call
        self.sum_attention_maps = None
        self.counter = 0


    def __call__(
        self,
        attn: Attention,
        hidden_states: torch.Tensor,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        temb: Optional[torch.Tensor] = None,
        *args,
        **kwargs,
    ) -> torch.Tensor:
        if len(args) > 0 or kwargs.get("scale", None) is not None:
            deprecation_message = "The `scale` argument is deprecated and will be ignored. Please remove it, as passing it will raise an error in the future. `scale` should directly be passed while calling the underlying pipeline component i.e., via `cross_attention_kwargs`."
            deprecate("scale", "1.0.0", deprecation_message)

        residual = hidden_states

        if attn.spatial_norm is not None:
            hidden_states = attn.spatial_norm(hidden_states, temb)

        input_ndim = hidden_states.ndim

        if input_ndim == 4:
            batch_size, channel, height, width = hidden_states.shape
            hidden_states = hidden_states.view(batch_size, channel, height * width).transpose(1, 2)

        batch_size, sequence_length, _ = (
            hidden_states.shape if encoder_hidden_states is None else encoder_hidden_states.shape
        )
        attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)

        if attn.group_norm is not None:
            hidden_states = attn.group_norm(hidden_states.transpose(1, 2)).transpose(1, 2)

        query = attn.to_q(hidden_states)

        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)

        key = attn.to_k(encoder_hidden_states)
        value = attn.to_v(encoder_hidden_states)

        query = attn.head_to_batch_dim(query)
        key = attn.head_to_batch_dim(key)
        value = attn.head_to_batch_dim(value)

        attention_probs = attn.get_attention_scores(query, key, attention_mask)

        res = int(np.sqrt(attention_probs.shape[1]))
        num_tokens = attention_probs.shape[-1]
        attention_map = attention_probs.reshape(-1, res, res, num_tokens)
        if self.sum_attention_maps is None:
            self.sum_attention_maps = attention_map
        else:
            self.sum_attention_maps += attention_map
        self.counter += 1

        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)
        # dropout
        hidden_states = attn.to_out[1](hidden_states)

        if input_ndim == 4:
            hidden_states = hidden_states.transpose(-1, -2).reshape(batch_size, channel, height, width)

        if attn.residual_connection:
            hidden_states = hidden_states + residual

        hidden_states = hidden_states / attn.rescale_output_factor

        return hidden_states