File size: 14,485 Bytes
ddf288e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
from fast_pytorch_kmeans import KMeans
from torch import einsum
import torch.distributed as dist
from einops import rearrange


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)

    def forward(self, x):
        x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
        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:
            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


class ResnetBlock(nn.Module):
    def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False, dropout):
        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.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:
                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):
        h = x
        h = self.norm1(h)
        h = nonlinearity(h)
        h = self.conv1(h)

        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)
        h_ = torch.bmm(v,w_)     # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
        h_ = h_.reshape(b,c,h,w)

        h_ = self.proj_out(h_)

        return x+h_


class Swish(nn.Module):
    def forward(self, x):
        return x * torch.sigmoid(x)


class Encoder(nn.Module):
    """
    Encoder of VQ-GAN to map input batch of images to latent space.
    Dimension Transformations:
    3x256x256 --Conv2d--> 32x256x256
    for loop:
              --ResBlock--> 64x256x256 --DownBlock--> 64x128x128
              --ResBlock--> 128x128x128 --DownBlock--> 128x64x64
              --ResBlock--> 256x64x64  --DownBlock--> 256x32x32
              --ResBlock--> 512x32x32
    --ResBlock--> 512x32x32
    --NonLocalBlock--> 512x32x32
    --ResBlock--> 512x32x32
    --GroupNorm-->
    --Swish-->
    --Conv2d-> 256x32x32
    """

    def __init__(self, in_channels=3, channels=[128, 128, 128, 256, 512, 512], attn_resolutions=[32], resolution=512, dropout=0.0, num_res_blocks=2, z_channels=256, **kwargs):
        super(Encoder, self).__init__()
        layers = [nn.Conv2d(in_channels, channels[0], 3, 1, 1)]
        for i in range(len(channels) - 1):
            in_channels = channels[i]
            out_channels = channels[i + 1]
            for j in range(num_res_blocks):
                layers.append(ResnetBlock(in_channels=in_channels, out_channels=out_channels, dropout=0.0))
                in_channels = out_channels
                if resolution in attn_resolutions:
                    layers.append(AttnBlock(in_channels))
            if i < len(channels) - 2:
                layers.append(Downsample(channels[i + 1], with_conv=True))
                resolution //= 2
        layers.append(ResnetBlock(in_channels=channels[-1], out_channels=channels[-1], dropout=0.0))
        layers.append(AttnBlock(channels[-1]))
        layers.append(ResnetBlock(in_channels=channels[-1], out_channels=channels[-1], dropout=0.0))
        layers.append(Normalize(channels[-1]))
        layers.append(Swish())
        layers.append(nn.Conv2d(channels[-1], z_channels, 3, 1, 1))
        self.model = nn.Sequential(*layers)

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


class Decoder(nn.Module):
    def __init__(self, out_channels=3, channels=[128, 128, 128, 256, 512, 512], attn_resolutions=[32], resolution=512, dropout=0.0, num_res_blocks=2, z_channels=256, **kwargs):
        super(Decoder, self).__init__()
        ch_mult = channels[1:]
        num_resolutions = len(ch_mult)
        block_in = ch_mult[num_resolutions - 1]
        curr_res = resolution// 2 ** (num_resolutions - 1)

        layers = [nn.Conv2d(z_channels, block_in, kernel_size=3, stride=1, padding=1),
                  ResnetBlock(in_channels=block_in, out_channels=block_in, dropout=0.0),
                  AttnBlock(block_in),
                  ResnetBlock(in_channels=block_in, out_channels=block_in, dropout=0.0)
                  ]

        for i in reversed(range(num_resolutions)):
            block_out = ch_mult[i]
            for i_block in range(num_res_blocks+1):
                layers.append(ResnetBlock(in_channels=block_in, out_channels=block_out, dropout=0.0))
                block_in = block_out
                if curr_res in attn_resolutions:
                    layers.append(AttnBlock(block_in))
            if i > 0:
                layers.append(Upsample(block_in, with_conv=True))
            curr_res = curr_res * 2

        layers.append(Normalize(block_in))
        layers.append(Swish())
        layers.append(nn.Conv2d(block_in, out_channels, kernel_size=3, stride=1, padding=1))

        self.model = nn.Sequential(*layers)

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


class Codebook(nn.Module):
    """
    Improved version over VectorQuantizer, can be used as a drop-in replacement. Mostly
    avoids costly matrix multiplications and allows for post-hoc remapping of indices.
    """
    def __init__(self, codebook_size, codebook_dim, beta, init_steps=2000, reservoir_size=2e5):
        super().__init__()
        self.codebook_size = codebook_size
        self.codebook_dim = codebook_dim
        self.beta = beta

        self.embedding = nn.Embedding(self.codebook_size, self.codebook_dim)
        self.embedding.weight.data.uniform_(-1.0 / self.codebook_size, 1.0 / self.codebook_size)

        self.q_start_collect, self.q_init, self.q_re_end, self.q_re_step = init_steps, init_steps * 3, init_steps * 30, init_steps // 2
        self.q_counter = 0
        self.reservoir_size = int(reservoir_size)
        self.reservoir = None

    def forward(self, z):
        z = rearrange(z, 'b c h w -> b h w c').contiguous()
        batch_size = z.size(0)
        z_flattened = z.view(-1, self.codebook_dim)
        if self.training:
            self.q_counter += 1
            # x_flat = x.permute(0, 2, 3, 1).reshape(-1, z.shape(1))
            if self.q_counter > self.q_start_collect:
                z_new = z_flattened.clone().detach().view(batch_size, -1, self.codebook_dim)
                z_new = z_new[:, torch.randperm(z_new.size(1))][:, :10].reshape(-1, self.codebook_dim)
                self.reservoir = z_new if self.reservoir is None else torch.cat([self.reservoir, z_new], dim=0)
                self.reservoir = self.reservoir[torch.randperm(self.reservoir.size(0))[:self.reservoir_size]].detach()
            if self.q_counter < self.q_init:
                z_q = rearrange(z, 'b h w c -> b c h w').contiguous()
                return z_q, z_q.new_tensor(0), None  # z_q, loss, min_encoding_indices
            else:
                # if self.q_counter < self.q_init + self.q_re_end:
                if self.q_init <= self.q_counter < self.q_re_end:
                    if (self.q_counter - self.q_init) % self.q_re_step == 0 or self.q_counter == self.q_init + self.q_re_end - 1:
                        kmeans = KMeans(n_clusters=self.codebook_size)
                        world_size = dist.get_world_size()
                        print("Updating codebook from reservoir.")
                        if world_size > 1:
                            global_reservoir = [torch.zeros_like(self.reservoir) for _ in range(world_size)]
                            dist.all_gather(global_reservoir, self.reservoir.clone())
                            global_reservoir = torch.cat(global_reservoir, dim=0)
                        else:
                            global_reservoir = self.reservoir
                        kmeans.fit_predict(global_reservoir)  # reservoir is 20k encoded latents
                        self.embedding.weight.data = kmeans.centroids.detach()

        d = torch.sum(z_flattened ** 2, dim=1, keepdim=True) + \
            torch.sum(self.embedding.weight**2, dim=1) - 2 * \
            torch.einsum('bd,dn->bn', z_flattened, rearrange(self.embedding.weight, 'n d -> d n'))

        min_encoding_indices = torch.argmin(d, dim=1)
        z_q = self.embedding(min_encoding_indices).view(z.shape)

        # compute loss for embedding
        loss = torch.mean((z_q.detach()-z)**2) + self.beta * torch.mean((z_q - z.detach()) ** 2)

        # preserve gradients
        z_q = z + (z_q - z).detach()

        # reshape back to match original input shape
        z_q = rearrange(z_q, 'b h w c -> b c h w').contiguous()

        return z_q, loss, min_encoding_indices

    def get_codebook_entry(self, indices, shape):
        # get quantized latent vectors
        z_q = self.embedding(indices)

        if shape is not None:
            z_q = z_q.view(shape)
            # reshape back to match original input shape
            z_q = z_q.permute(0, 3, 1, 2).contiguous()

        return z_q


if __name__ == '__main__':
    enc = Encoder()
    dec = Decoder()
    print(sum([p.numel() for p in enc.parameters()]))
    print(sum([p.numel() for p in dec.parameters()]))
    x = torch.randn(1, 3, 512, 512)
    res = enc(x)
    print(res.shape)
    res = dec(res)
    print(res.shape)