File size: 13,862 Bytes
5359669
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
# This script is from the following repositories
# https://github.com/ermongroup/ddim
# https://github.com/bahjat-kawar/ddrm


def get_timestep_embedding(timesteps, embedding_dim):
    """
    This matches the implementation in Denoising Diffusion Probabilistic Models:
    From Fairseq.
    Build sinusoidal embeddings.
    This matches the implementation in tensor2tensor, but differs slightly
    from the description in Section 3.5 of "Attention Is All You Need".
    """
    assert len(timesteps.shape) == 1

    half_dim = embedding_dim // 2
    emb = math.log(10000) / (half_dim - 1)
    emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
    emb = emb.to(device=timesteps.device)
    emb = timesteps.float()[:, None] * emb[None, :]
    emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
    if embedding_dim % 2 == 1:  # zero pad
        emb = torch.nn.functional.pad(emb, (0, 1, 0, 0))
    return emb


def nonlinearity(x):
    # swish
    return x*torch.sigmoid(x)


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


class Upsample(nn.Module):
    def __init__(self, in_channels, with_conv):
        super().__init__()
        self.with_conv = with_conv
        if self.with_conv:
            self.conv = torch.nn.Conv2d(in_channels,in_channels,kernel_size=3,stride=1,padding=1)
            #padding=1. The convoution won't change the input shape 

    def forward(self, x):
        x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
        #scale_factor. Perform up sampling to the input. The image size becomes (image_h*2) * (image_w*2)
        if self.with_conv:
            x = self.conv(x)
        return x


class Downsample(nn.Module):
    def __init__(self, in_channels, with_conv):
        super().__init__()
        self.with_conv = with_conv
        if self.with_conv:
            # no asymmetric padding in torch conv, must do it ourselves
            self.conv = torch.nn.Conv2d(in_channels,in_channels,kernel_size=3,stride=2,padding=0)

    def forward(self, x):
        if self.with_conv:#whether use convoution to perform downsampling
            pad = (0, 1, 0, 1)
            x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
            x = self.conv(x)
        else:
            x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)
        return x
        #This is down sampling, The image size becomes image_h/2 * image_w/2


class ResnetBlock(nn.Module):
    def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False,
                 dropout, temb_channels=512):
        super().__init__()
        self.in_channels = in_channels
        out_channels = in_channels if out_channels is None else out_channels
        self.out_channels = out_channels
        self.use_conv_shortcut = conv_shortcut

        self.norm1 = Normalize(in_channels)
        self.conv1 = torch.nn.Conv2d(in_channels,out_channels,kernel_size=3,stride=1,padding=1)
        self.temb_proj = torch.nn.Linear(temb_channels,out_channels)#projection
        self.norm2 = Normalize(out_channels)
        self.dropout = torch.nn.Dropout(dropout)
        self.conv2 = torch.nn.Conv2d(out_channels,out_channels,kernel_size=3,stride=1,padding=1)
        if self.in_channels != self.out_channels:
            if self.use_conv_shortcut:#whether use convoution to perform residual
                self.conv_shortcut = torch.nn.Conv2d(in_channels,out_channels,kernel_size=3,stride=1,padding=1)
            else:
                self.nin_shortcut = torch.nn.Conv2d(in_channels,out_channels,kernel_size=1,stride=1,padding=0)

    def forward(self, x, temb):
        h = x
        h = self.norm1(h)
        h = nonlinearity(h)
        h = self.conv1(h)
        #temb=token embedding dimension
        h = h + self.temb_proj(nonlinearity(temb))[:, :, None, None]

        h = self.norm2(h)
        h = nonlinearity(h)
        h = self.dropout(h)
        h = self.conv2(h)

        if self.in_channels != self.out_channels:
            if self.use_conv_shortcut:
                x = self.conv_shortcut(x)
            else:
                x = self.nin_shortcut(x)

        return x+h


class AttnBlock(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 = q.reshape(b, c, h*w)
        q = q.permute(0, 2, 1)   # b,hw,c
        k = k.reshape(b, c, h*w)  # b,c,hw
        w_ = torch.bmm(q, k)     # b,hw,hw    w[b,i,j]=sum_c q[b,i,c]k[b,c,j]
        w_ = w_ * (int(c)**(-0.5))
        w_ = torch.nn.functional.softmax(w_, dim=2)

        # attend to values
        v = v.reshape(b, c, h*w)
        w_ = w_.permute(0, 2, 1)   # b,hw,hw (first hw of k, second of q)
        # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
        h_ = torch.bmm(v, w_)
        h_ = h_.reshape(b, c, h, w)

        h_ = self.proj_out(h_)

        return x+h_

class MultiHeadAttnBlock(nn.Module):
    def __init__(self, in_channels):
        super().__init__()
        num_heads = 8
        assert in_channels % num_heads == 0, "in_channels must be divisible by num_heads"
        
        self.in_channels = in_channels
        self.num_heads = num_heads
        self.head_dim = in_channels // num_heads

        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_)

        # Split heads
        b, c, h, w = q.shape
        q = q.reshape(b, self.num_heads, self.head_dim, h * w).permute(0, 1, 3, 2)  # b, num_heads, hw, head_dim
        k = k.reshape(b, self.num_heads, self.head_dim, h * w)  # b, num_heads, head_dim, hw
        v = v.reshape(b, self.num_heads, self.head_dim, h * w)  # b, num_heads, head_dim, hw

        # Compute attention
        w_ = torch.einsum('bnqd,bnkd->bnqk', q, k)  # b, num_heads, hw, hw
        w_ = w_ * (self.head_dim ** -0.5)
        w_ = torch.nn.functional.softmax(w_, dim=-1)

        # Attend to values
        h_ = torch.einsum('bnqk,bnvd->bnqd', w_, v)
        h_ = h_.permute(0, 1, 3, 2).reshape(b, c, h, w)  # Merge heads

        h_ = self.proj_out(h_)

        return x + h_
        
class DiffusionUNet(nn.Module):
    def __init__(self,ch, num_res_blocks, image_size, drop_out):
        super().__init__()
        self.dropout=drop_out

        ch_mult = [1, 1, 2, 2, 4, 4]
        
        attn_resolutions = [32,16,]
        #only when the image becomes attn_resolutions*attn_resolutions, use the self attention.
        resamp_with_conv = True
        #dropout = config.model.dropout
        #in_channels = config.model.in_channels * 2 if config.data.conditional else config.model.in_channels
        in_channels = 2
        out_ch = 1
        resolution = image_size
        
        self.ch = ch
        self.temb_ch = self.ch*4 #the time step embedding dimension
        self.num_resolutions = len(ch_mult) #the nubmer of down-sampling and up-sampling
        #the ch_mult is a tuple, specify the channels in different level of feature extraction (down-sampling or sampling)
        self.num_res_blocks = num_res_blocks
        self.resolution = resolution
        self.in_channels = in_channels
        
        #self.shearemb = nn.Module()
        #self.shearemb.dense = nn.ModuleList([
        #    torch.nn.Linear(3, self.temb_ch),
        #    torch.nn.Linear(self.temb_ch, self.temb_ch),])
        self.shear_emb = nn.Embedding(num_embeddings=3, embedding_dim=self.temb_ch)
        
        # timestep embedding
        self.temb = nn.Module()
        self.temb.dense = nn.ModuleList([
            torch.nn.Linear(self.ch, self.temb_ch),
            torch.nn.Linear(self.temb_ch, self.temb_ch),])

        # downsampling
        self.conv_in = torch.nn.Conv2d(in_channels, self.ch, kernel_size = 3, stride = 1, padding = 1) #the first feature extraction
        
        curr_res = resolution
        in_ch_mult = (1,)+tuple(ch_mult)
        self.down = nn.ModuleList()
        block_in = None
        for i_level in range(self.num_resolutions):
            block = nn.ModuleList()
            attn = nn.ModuleList()
            block_in = ch*in_ch_mult[i_level]
            block_out = ch*ch_mult[i_level]
            for i_block in range(self.num_res_blocks):
                block.append(ResnetBlock(in_channels = block_in, out_channels = block_out, temb_channels = self.temb_ch, dropout = self.dropout))
                block_in = block_out
                if curr_res in attn_resolutions:
                    attn.append(AttnBlock(block_in))
            down = nn.Module()
            down.block = block
            down.attn = attn
            if i_level != self.num_resolutions-1:
                down.downsample = Downsample(block_in, resamp_with_conv)
                curr_res = curr_res // 2 #afte one down-sampling, the resolution becomes half.
            self.down.append(down)

        # middle
        self.mid = nn.Module()
        self.mid.block_1 = ResnetBlock(in_channels=block_in,out_channels=block_in,temb_channels=self.temb_ch,dropout=self.dropout)
        self.mid.attn_1 = AttnBlock(block_in)
        self.mid.block_2 = ResnetBlock(in_channels=block_in,out_channels=block_in,temb_channels=self.temb_ch,dropout=self.dropout)

        # upsampling
        self.up = nn.ModuleList()
        for i_level in reversed(range(self.num_resolutions)):
            block = nn.ModuleList()
            attn = nn.ModuleList()
            block_out = ch*ch_mult[i_level]
            skip_in = ch*ch_mult[i_level]
            for i_block in range(self.num_res_blocks+1):
                if i_block == self.num_res_blocks:
                    skip_in = ch*in_ch_mult[i_level]
                block.append(ResnetBlock(in_channels=block_in+skip_in,out_channels=block_out,temb_channels=self.temb_ch,dropout=self.dropout))
                block_in = block_out
                if curr_res in attn_resolutions:
                    attn.append(AttnBlock(block_in))
            up = nn.Module()
            up.block = block
            up.attn = attn
            if i_level != 0:
                up.upsample = Upsample(block_in, resamp_with_conv)
                curr_res = curr_res * 2 #after one up-sampling, the resolution becomes twice.
            self.up.insert(0, up)  # prepend to get consistent order

        # end
        self.norm_out = Normalize(block_in)
        self.conv_out = torch.nn.Conv2d(block_in,out_ch,kernel_size=3,stride=1,padding=1)
        
    def forward(self, x, t, shear):
        #condition is the initial condition used as key and value in cross attention
        
        assert x.shape[2] == x.shape[3] == self.resolution
        #print(x.shape)
        # timestep embedding
        temb = get_timestep_embedding(t, self.ch)
        temb = self.temb.dense[0](temb)
        temb = nonlinearity(temb)
        temb = self.temb.dense[1](temb)

        #shear embedding
        shear_emb = self.shear_emb(shear)
        '''
        shear_emb=nn.functional.one_hot(shear, num_classes=3).type(torch.float) #one hot
        shear_emb=self.shearemb.dense[0](shear_emb)
        shear_emb = nonlinearity(shear_emb)
        shear_emb=self.shearemb.dense[1](shear_emb)
        '''

        temb = temb + shear_emb
        
        # downsampling
        hs = [self.conv_in(x)] #the first feature extraction
        for i_level in range(self.num_resolutions):
            for i_block in range(self.num_res_blocks):
                h = self.down[i_level].block[i_block](hs[-1], temb) #perform one Restblock
                if len(self.down[i_level].attn) > 0: #check if there is Attblock
                    h = self.down[i_level].attn[i_block](h)
                hs.append(h)
            if i_level != self.num_resolutions-1:
                hs.append(self.down[i_level].downsample(hs[-1])) #perform down-sampling
            #store the each level down-sampling ouput for skip connection when up-sampling

        # middle
        h = hs[-1]
        h = self.mid.block_1(h, temb)
        h = self.mid.attn_1(h)
        h = self.mid.block_2(h, temb)

        # upsampling
        for i_level in reversed(range(self.num_resolutions)):
            for i_block in range(self.num_res_blocks+1):
                h = self.up[i_level].block[i_block](
                    torch.cat([h, hs.pop()], dim=1), temb) #this is skip connection
                if len(self.up[i_level].attn) > 0:
                    h = self.up[i_level].attn[i_block](h)
            if i_level != 0:
                h = self.up[i_level].upsample(h)

        # end
        h = self.norm_out(h)
        h = nonlinearity(h)
        h = self.conv_out(h)
        return h