ldm/modules/attention.py

from inspect import isfunction
import math
import torch
import torch.nn.functional as F
from torch import nn, einsum
from einops import rearrange, repeat
from typing import Optional, Any
import matplotlib.pyplot as plt
import numpy as np
from ldm.modules.diffusionmodules.util import checkpoint
from PIL import Image

try:
    import xformers
    import xformers.ops
    XFORMERS_IS_AVAILBLE = False
except:
    XFORMERS_IS_AVAILBLE = False


def exists(val):
    return val is not None


def uniq(arr):
    return{el: True for el in arr}.keys()


def default(val, d):
    if exists(val):
        return val
    return d() if isfunction(d) else d


def max_neg_value(t):
    return -torch.finfo(t.dtype).max


def init_(tensor):
    dim = tensor.shape[-1]
    std = 1 / math.sqrt(dim)
    tensor.uniform_(-std, std)
    return tensor


# feedforward
class GEGLU(nn.Module):
    def __init__(self, dim_in, dim_out):
        super().__init__()
        self.proj = nn.Linear(dim_in, dim_out * 2)

    def forward(self, x):
        x, gate = self.proj(x).chunk(2, dim=-1)
        return x * F.gelu(gate)


class FeedForward(nn.Module):
    def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
        super().__init__()
        inner_dim = int(dim * mult)
        dim_out = default(dim_out, dim)
        project_in = nn.Sequential(
            nn.Linear(dim, inner_dim),
            nn.GELU()
        ) if not glu else GEGLU(dim, inner_dim)

        self.net = nn.Sequential(
            project_in,
            nn.Dropout(dropout),
            nn.Linear(inner_dim, dim_out)
        )

    def forward(self, x):
        return self.net(x)


def zero_module(module):
    """
    Zero out the parameters of a module and return it.
    """
    for p in module.parameters():
        p.detach().zero_()
    return module


def Normalize(in_channels):
    return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)


class SpatialSelfAttention(nn.Module):
    def __init__(self, in_channels):
        super().__init__()
        self.in_channels = in_channels

        self.norm = Normalize(in_channels)
        self.q = torch.nn.Conv2d(in_channels,
                                 in_channels,
                                 kernel_size=1,
                                 stride=1,
                                 padding=0)
        self.k = torch.nn.Conv2d(in_channels,
                                 in_channels,
                                 kernel_size=1,
                                 stride=1,
                                 padding=0)
        self.v = torch.nn.Conv2d(in_channels,
                                 in_channels,
                                 kernel_size=1,
                                 stride=1,
                                 padding=0)
        self.proj_out = torch.nn.Conv2d(in_channels,
                                        in_channels,
                                        kernel_size=1,
                                        stride=1,
                                        padding=0)

    def forward(self, x):
        h_ = x
        h_ = self.norm(h_)
        q = self.q(h_)
        k = self.k(h_)
        v = self.v(h_)

        # compute attention
        b,c,h,w = q.shape
        q = rearrange(q, 'b c h w -> b (h w) c')
        k = rearrange(k, 'b c h w -> b c (h w)')
        w_ = torch.einsum('bij,bjk->bik', q, k)

        w_ = w_ * (int(c)**(-0.5))
        w_ = torch.nn.functional.softmax(w_, dim=2)

        # attend to values
        v = rearrange(v, 'b c h w -> b c (h w)')
        w_ = rearrange(w_, 'b i j -> b j i')
        h_ = torch.einsum('bij,bjk->bik', v, w_)
        h_ = rearrange(h_, 'b c (h w) -> b c h w', h=h)
        h_ = self.proj_out(h_)

        return x+h_


class CrossAttention(nn.Module):
    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.):
        super().__init__()
        inner_dim = dim_head * heads
        context_dim = default(context_dim, query_dim)

        self.scale = dim_head ** -0.5
        self.heads = heads

        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)

        self.to_out = nn.Sequential(
            nn.Linear(inner_dim, query_dim),
            nn.Dropout(dropout)
        )

    def forward(self, x, context=None, mask=None, encode=False, controller_for_inject=None, inject=False, layernum=None, main_height=None, main_width=None):
        h = self.heads

        q = self.to_q(x)
        context = default(context, x)
        k = self.to_k(context)
        v = self.to_v(context)

        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))

        sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
        del q, k

        if exists(mask):
            mask = rearrange(mask, 'b ... -> b (...)')
            max_neg_value = -torch.finfo(sim.dtype).max
            mask = repeat(mask, 'b j -> (b h) () j', h=h)
            sim.masked_fill_(~mask, max_neg_value)

        # a = ((sim.mean(0).mean(1).resize(64,64)/torch.max(sim.max(), abs(sim.min())) + 1)*127.5).cpu().numpy().astype(np.uint8)
        # image = Image.fromarray(a)
        # image.resize((512,512)).save('2.jpg')
        
        # u, s, vh = np.linalg.svd(sim.mean(0).cpu().numpy().astype(np.float32) - np.mean(sim.mean(0).cpu().numpy().astype(np.float32), axis=1, keepdims=True))
        # images = []
        # for i in range(3):
        #     image = u[:,i].reshape(64, 64)
        #     image = image - image.min()
        #     image = 255 * image / image.max()
        #     image = np.expand_dims(image, axis=2).astype(np.uint8)
        #     images.append(image)

        # final = np.dstack(images)   
        # final = Image.fromarray(final).resize((256, 256))
        # final = np.array(final)   
        # import ptp_scripts.ptp_utils as ptp_utils
        # ptp_utils.view_images(final)
                
        # attention, what we cannot get enough of
        sim = sim.softmax(dim=-1)
        # sim = sim.sigmoid()
        
        out = einsum('b i j, b j d -> b i d', sim, v)
        out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
        return self.to_out(out)


class MemoryEfficientCrossAttention(nn.Module):
    # https://github.com/MatthieuTPHR/diffusers/blob/d80b531ff8060ec1ea982b65a1b8df70f73aa67c/src/diffusers/models/attention.py#L223
    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.0):
        super().__init__()
        print(f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using "
              f"{heads} heads.")
        inner_dim = dim_head * heads
        context_dim = default(context_dim, query_dim)

        self.heads = heads
        self.dim_head = dim_head

        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)

        self.to_out = nn.Sequential(nn.Linear(inner_dim, query_dim), nn.Dropout(dropout))
        self.attention_op: Optional[Any] = None

    def forward(self, x, context=None, mask=None):
        q = self.to_q(x)
        context = default(context, x)
        k = self.to_k(context)
        v = self.to_v(context)

        b, _, _ = q.shape
        q, k, v = map(
            lambda t: t.unsqueeze(3)
            .reshape(b, t.shape[1], self.heads, self.dim_head)
            .permute(0, 2, 1, 3)
            .reshape(b * self.heads, t.shape[1], self.dim_head)
            .contiguous(),
            (q, k, v),
        )

        # actually compute the attention, what we cannot get enough of
        out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)

        if exists(mask):
            raise NotImplementedError
        out = (
            out.unsqueeze(0)
            .reshape(b, self.heads, out.shape[1], self.dim_head)
            .permute(0, 2, 1, 3)
            .reshape(b, out.shape[1], self.heads * self.dim_head)
        )
        return self.to_out(out)


class BasicTransformerBlock(nn.Module):
    ATTENTION_MODES = {
        "softmax": CrossAttention,  # vanilla attention
        "softmax-xformers": MemoryEfficientCrossAttention
    }
    def __init__(self, dim, n_heads, d_head, dropout=0., context_dim=None, gated_ff=True, checkpoint=True,
                 disable_self_attn=False):
        super().__init__()
        attn_mode = "softmax-xformers" if XFORMERS_IS_AVAILBLE else "softmax"
        assert attn_mode in self.ATTENTION_MODES
        attn_cls = self.ATTENTION_MODES[attn_mode]
        self.disable_self_attn = disable_self_attn
        self.attn1 = attn_cls(query_dim=dim, heads=n_heads, dim_head=d_head, dropout=dropout,
                              context_dim=context_dim if self.disable_self_attn else None)  # is a self-attention if not self.disable_self_attn
        self.ff = FeedForward(dim, dropout=dropout, glu=gated_ff)
        self.attn2 = attn_cls(query_dim=dim, context_dim=context_dim,
                              heads=n_heads, dim_head=d_head, dropout=dropout)  # is self-attn if context is none
        self.norm1 = nn.LayerNorm(dim)
        self.norm2 = nn.LayerNorm(dim)
        self.norm3 = nn.LayerNorm(dim)
        self.checkpoint = checkpoint

    def forward(self, x, context=None, encode=False, encode_uncon=False, decode_uncon=False, controller=None, inject=False, layernum=0, h=None, w=None):
        return checkpoint(self._forward, (x, context, encode, encode_uncon, decode_uncon, controller, inject, layernum, h, w), self.parameters(), self.checkpoint)

    def _forward(self, x, context=None, encode=False, encode_uncon=False, decode_uncon=False, controller=None, inject=False, layernum=0, h=None, w=None):
        
        if encode_uncon == True and decode_uncon == True:
            # pass
            x = self.attn1(self.norm1(x), context=None, encode=encode) + x
            x = self.attn1(self.norm1(x), context=None, encode=encode) + x # 如果要加层数，要记得改 register_attention_control
            
        elif encode_uncon == True and decode_uncon == False:
            if encode:
                x = self.attn1(self.norm1(x), context=None, encode=encode) + x
                x = self.attn1(self.norm1(x), context=None, encode=encode) + x # 如果要加层数，要记得改 register_attention_control
            else:
                x = self.attn1(self.norm1(x), context=context 
                               if self.disable_self_attn else None, controller_for_inject=controller, inject=inject, layernum=layernum) + x
                x = self.attn1(self.norm1(x), context=context 
                               if self.disable_self_attn else None, controller_for_inject=controller, inject=inject, layernum=layernum+1) + x
                x = self.attn2(self.norm2(x), context=context) + x
        
        elif encode_uncon == False and decode_uncon == False:
            x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None, encode=encode, 
                           controller_for_inject=controller, inject=inject, layernum=layernum, main_height=h, main_width=w) + x
            x = self.attn2(self.norm2(x), context=context, encode=encode) + x
            # pass
            
        x = self.ff(self.norm3(x)) + x
        return x


class SpatialTransformer(nn.Module):
    """
    Transformer block for image-like data.
    First, project the input (aka embedding)
    and reshape to b, t, d.
    Then apply standard transformer action.
    Finally, reshape to image
    NEW: use_linear for more efficiency instead of the 1x1 convs
    """
    def __init__(self, in_channels, n_heads, d_head,
                 depth=1, dropout=0., context_dim=None,
                 disable_self_attn=False, use_linear=False,
                 use_checkpoint=True):
        super().__init__()
        if exists(context_dim) and not isinstance(context_dim, list):
            context_dim = [context_dim]
        self.in_channels = in_channels
        inner_dim = n_heads * d_head
        self.norm = Normalize(in_channels)
        if not use_linear:
            self.proj_in = nn.Conv2d(in_channels,
                                     inner_dim,
                                     kernel_size=1,
                                     stride=1,
                                     padding=0)
        else:
            self.proj_in = nn.Linear(in_channels, inner_dim)

        self.transformer_blocks = nn.ModuleList(
            [BasicTransformerBlock(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim[d],
                                   disable_self_attn=disable_self_attn, checkpoint=use_checkpoint)
                for d in range(depth)]
        )
        if not use_linear:
            self.proj_out = zero_module(nn.Conv2d(inner_dim,
                                                  in_channels,
                                                  kernel_size=1,
                                                  stride=1,
                                                  padding=0))
        else:
            self.proj_out = zero_module(nn.Linear(in_channels, inner_dim))
        self.use_linear = use_linear

    def forward(self, x, context=None, encode=False, encode_uncon=True, decode_uncon=True, controller=None, inject=False, layernum=0):
        # note: if no context is given, cross-attention defaults to self-attention
        if not isinstance(context, list):
            context = [context]
        b, c, h, w = x.shape
        x_in = x
        x = self.norm(x)
        if not self.use_linear:
            x = self.proj_in(x)
        x = rearrange(x, 'b c h w -> b (h w) c').contiguous()
        if self.use_linear:
            x = self.proj_in(x)
        for i, block in enumerate(self.transformer_blocks):
            x = block(x, context=context[i], encode=encode, encode_uncon=encode_uncon, decode_uncon=decode_uncon, 
                      controller=controller, inject=inject, layernum=layernum, h=h, w=w)
        if self.use_linear:
            x = self.proj_out(x)
        x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w).contiguous()
        if not self.use_linear:
            x = self.proj_out(x)
        
        layernum = layernum + 1 # 和register_recr对应起来
        
        return x + x_in, layernum