sa_handler.py

# Copyright 2023 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.


from __future__ import annotations

from dataclasses import dataclass
from diffusers import StableDiffusionXLPipeline
import torch
import torch.nn as nn
from torch.nn import functional as nnf
from diffusers.models import attention_processor
import einops

T = torch.Tensor


@dataclass(frozen=True)
class StyleAlignedArgs:
    share_group_norm: bool = True
    share_layer_norm: bool = True,
    share_attention: bool = True
    adain_queries: bool = True
    adain_keys: bool = True
    adain_values: bool = False
    full_attention_share: bool = False
    shared_score_scale: float = 1.
    shared_score_shift: float = 0.
    only_self_level: float = 0.


def expand_first(feat: T, scale=1.,) -> T:
    b = feat.shape[0]
    feat_style = torch.stack((feat[0], feat[b // 2])).unsqueeze(1)
    if scale == 1:
        feat_style = feat_style.expand(2, b // 2, *feat.shape[1:])
    else:
        feat_style = feat_style.repeat(1, b // 2, 1, 1, 1)
        feat_style = torch.cat([feat_style[:, :1], scale * feat_style[:, 1:]], dim=1)
    return feat_style.reshape(*feat.shape)


def concat_first(feat: T, dim=2, scale=1.) -> T:
    feat_style = expand_first(feat, scale=scale)
    return torch.cat((feat, feat_style), dim=dim)


def calc_mean_std(feat, eps: float = 1e-5) -> tuple[T, T]:
    feat_std = (feat.var(dim=-2, keepdims=True) + eps).sqrt()
    feat_mean = feat.mean(dim=-2, keepdims=True)
    return feat_mean, feat_std


def adain(feat: T) -> T:
    feat_mean, feat_std = calc_mean_std(feat)
    feat_style_mean = expand_first(feat_mean)
    feat_style_std = expand_first(feat_std)
    feat = (feat - feat_mean) / feat_std
    feat = feat * feat_style_std + feat_style_mean
    return feat


class DefaultAttentionProcessor(nn.Module):

    def __init__(self):
        super().__init__()
        self.processor = attention_processor.AttnProcessor2_0()

    def __call__(self, attn: attention_processor.Attention, hidden_states, encoder_hidden_states=None,
                 attention_mask=None, **kwargs):
        return self.processor(attn, hidden_states, encoder_hidden_states, attention_mask)


class SharedAttentionProcessor(DefaultAttentionProcessor):

    def shifted_scaled_dot_product_attention(self, attn: attention_processor.Attention, query: T, key: T, value: T) -> T:
        logits = torch.einsum('bhqd,bhkd->bhqk', query, key) * attn.scale
        logits[:, :, :, query.shape[2]:] += self.shared_score_shift
        probs = logits.softmax(-1)
        return torch.einsum('bhqk,bhkd->bhqd', probs, value)

    def shared_call(
            self,
            attn: attention_processor.Attention,
            hidden_states,
            encoder_hidden_states=None,
            attention_mask=None,
            **kwargs
    ):

        residual = hidden_states
        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
        )

        if attention_mask is not None:
            attention_mask = attn.prepare_attention_mask(attention_mask, sequence_length, batch_size)
            # scaled_dot_product_attention expects attention_mask shape to be
            # (batch, heads, source_length, target_length)
            attention_mask = attention_mask.view(batch_size, attn.heads, -1, attention_mask.shape[-1])

        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)
        key = attn.to_k(hidden_states)
        value = attn.to_v(hidden_states)
        inner_dim = key.shape[-1]
        head_dim = inner_dim // attn.heads

        query = query.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
        key = key.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
        value = value.view(batch_size, -1, attn.heads, head_dim).transpose(1, 2)
        # if self.step >= self.start_inject:
        if self.adain_queries:
            query = adain(query)
        if self.adain_keys:
            key = adain(key)
        if self.adain_values:
            value = adain(value)
        if self.share_attention:
            key = concat_first(key, -2, scale=self.shared_score_scale)
            value = concat_first(value, -2)
            if self.shared_score_shift != 0:
                hidden_states = self.shifted_scaled_dot_product_attention(attn, query, key, value,)
            else:
                hidden_states = nnf.scaled_dot_product_attention(
                    query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
                )
        else:
            hidden_states = nnf.scaled_dot_product_attention(
                query, key, value, attn_mask=attention_mask, dropout_p=0.0, is_causal=False
            )
        # hidden_states = adain(hidden_states)
        hidden_states = hidden_states.transpose(1, 2).reshape(batch_size, -1, attn.heads * head_dim)
        hidden_states = hidden_states.to(query.dtype)

        # 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

    def __call__(self, attn: attention_processor.Attention, hidden_states, encoder_hidden_states=None,
                 attention_mask=None, **kwargs):
        if self.full_attention_share:
            b, n, d = hidden_states.shape
            hidden_states = einops.rearrange(hidden_states, '(k b) n d -> k (b n) d', k=2)
            hidden_states = super().__call__(attn, hidden_states, encoder_hidden_states=encoder_hidden_states,
                                             attention_mask=attention_mask, **kwargs)
            hidden_states = einops.rearrange(hidden_states, 'k (b n) d -> (k b) n d', n=n)
        else:
            hidden_states = self.shared_call(attn, hidden_states, hidden_states, attention_mask, **kwargs)

        return hidden_states

    def __init__(self, style_aligned_args: StyleAlignedArgs):
        super().__init__()
        self.share_attention = style_aligned_args.share_attention
        self.adain_queries = style_aligned_args.adain_queries
        self.adain_keys = style_aligned_args.adain_keys
        self.adain_values = style_aligned_args.adain_values
        self.full_attention_share = style_aligned_args.full_attention_share
        self.shared_score_scale = style_aligned_args.shared_score_scale
        self.shared_score_shift = style_aligned_args.shared_score_shift


def _get_switch_vec(total_num_layers, level):
    if level == 0:
        return torch.zeros(total_num_layers, dtype=torch.bool)
    if level == 1:
        return torch.ones(total_num_layers, dtype=torch.bool)
    to_flip = level > .5
    if to_flip:
        level = 1 - level
    num_switch = int(level * total_num_layers)
    vec = torch.arange(total_num_layers)
    vec = vec % (total_num_layers // num_switch)
    vec = vec == 0
    if to_flip:
        vec = ~vec
    return vec


def init_attention_processors(pipeline: StableDiffusionXLPipeline, style_aligned_args: StyleAlignedArgs | None = None):
    attn_procs = {}
    unet = pipeline.unet
    number_of_self, number_of_cross = 0, 0
    num_self_layers = len([name for name in unet.attn_processors.keys() if 'attn1' in name])
    if style_aligned_args is None:
        only_self_vec = _get_switch_vec(num_self_layers, 1)
    else:
        only_self_vec = _get_switch_vec(num_self_layers, style_aligned_args.only_self_level)
    for i, name in enumerate(unet.attn_processors.keys()):
        is_self_attention = 'attn1' in name
        if is_self_attention:
            number_of_self += 1
            if style_aligned_args is None or only_self_vec[i // 2]:
                attn_procs[name] = DefaultAttentionProcessor()
            else:
                attn_procs[name] = SharedAttentionProcessor(style_aligned_args)
        else:
            number_of_cross += 1
            attn_procs[name] = DefaultAttentionProcessor()

    unet.set_attn_processor(attn_procs)


def register_shared_norm(pipeline: StableDiffusionXLPipeline,
                         share_group_norm: bool = True,
                         share_layer_norm: bool = True, ):
    def register_norm_forward(norm_layer: nn.GroupNorm | nn.LayerNorm) -> nn.GroupNorm | nn.LayerNorm:
        if not hasattr(norm_layer, 'orig_forward'):
            setattr(norm_layer, 'orig_forward', norm_layer.forward)
        orig_forward = norm_layer.orig_forward

        def forward_(hidden_states: T) -> T:
            n = hidden_states.shape[-2]
            hidden_states = concat_first(hidden_states, dim=-2)
            hidden_states = orig_forward(hidden_states)
            return hidden_states[..., :n, :]

        norm_layer.forward = forward_
        return norm_layer

    def get_norm_layers(pipeline_, norm_layers_: dict[str, list[nn.GroupNorm | nn.LayerNorm]]):
        if isinstance(pipeline_, nn.LayerNorm) and share_layer_norm:
            norm_layers_['layer'].append(pipeline_)
        if isinstance(pipeline_, nn.GroupNorm) and share_group_norm:
            norm_layers_['group'].append(pipeline_)
        else:
            for layer in pipeline_.children():
                get_norm_layers(layer, norm_layers_)

    norm_layers = {'group': [], 'layer': []}
    get_norm_layers(pipeline.unet, norm_layers)
    return [register_norm_forward(layer) for layer in norm_layers['group']] + [register_norm_forward(layer) for layer in
                                                                               norm_layers['layer']]


class Handler:

    def register(self, style_aligned_args: StyleAlignedArgs, ):
        self.norm_layers = register_shared_norm(self.pipeline, style_aligned_args.share_group_norm,
                                                style_aligned_args.share_layer_norm)
        init_attention_processors(self.pipeline, style_aligned_args)

    def remove(self):
        for layer in self.norm_layers:
            layer.forward = layer.orig_forward
        self.norm_layers = []
        init_attention_processors(self.pipeline, None)

    def __init__(self, pipeline: StableDiffusionXLPipeline):
        self.pipeline = pipeline
        self.norm_layers = []