File size: 12,298 Bytes
1bf3c29
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import numpy as np
import torch
import torch.nn as nn
from typing import Optional

from .modules import (
    ClassEmbedder,
    FinalLayer,
    FeedForward,
    MLP,
    RMSNorm,
    RotaryAttention,
    TimestepConditioner,
    apply_adaln,
    get_2d_sincos_pos_embed,
    get_2d_sincos_pos_embed_from_grid,
    precompute_freqs_cis_2d,
)


class PatchTokenEmbedder(nn.Module):
    def __init__(
            self,
            in_chans: int = 3,
            embed_dim: int = 768,
            norm_layer = None,
            bias: bool = True,
    ):
        super().__init__()
        self.in_chans = in_chans
        self.embed_dim = embed_dim
        self.proj = nn.Linear(in_chans, embed_dim, bias=bias)
        self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()

    def forward(self, x):
        x = self.proj(x)
        x = self.norm(x)
        return x


class AugmentedDiTBlock(nn.Module):
    def __init__(self, hidden_size, groups, mlp_ratio=4.0, adaLN_modulation=None):
        super().__init__()
        self.norm1 = RMSNorm(hidden_size, eps=1e-6)
        self.attn = RotaryAttention(hidden_size, num_heads=groups, qkv_bias=False)
        self.norm2 = RMSNorm(hidden_size, eps=1e-6)
        mlp_hidden_dim = int(hidden_size * mlp_ratio)
        self.mlp = FeedForward(hidden_size, mlp_hidden_dim)
        self.adaLN_modulation = adaLN_modulation if adaLN_modulation is not None else nn.Sequential(
            nn.Linear(hidden_size, 6 * hidden_size, bias=True)
        )

    def forward(self, x, c, pos, mask=None):
        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = self.adaLN_modulation(c).chunk(6, dim=-1)
        x = x + gate_msa * self.attn(apply_adaln(self.norm1(x), shift_msa, scale_msa), pos, mask=mask)
        x = x + gate_mlp * self.mlp(apply_adaln(self.norm2(x), shift_mlp, scale_mlp))
        return x


class PixelTokenEmbedder(nn.Module):
    def __init__(self, in_channels: int, hidden_size_output: int, use_pixel_abs_pos: bool = True):
        super().__init__()
        self.in_channels = int(in_channels)
        self.hidden_size_output = int(hidden_size_output)
        self.use_pixel_abs_pos = bool(use_pixel_abs_pos)
        self.proj = nn.Linear(self.in_channels, self.hidden_size_output, bias=True)
        self._pos_cache = dict()

    def _fetch_pixel_pos_image(self, height: int, width: int, device, dtype):
        if height == width:
            key = ("image", height, width)
            if key in self._pos_cache:
                pe = self._pos_cache[key]
                return pe.to(device=device, dtype=dtype)
            pos_np = get_2d_sincos_pos_embed(self.hidden_size_output, height)
            pos = torch.from_numpy(pos_np).to(device=device, dtype=dtype)
            self._pos_cache[key] = pos
            return pos
        else:
            key = ("image", height, width)
            if key in self._pos_cache:
                pe = self._pos_cache[key]
                return pe.to(device=device, dtype=dtype)
            grid_h = np.arange(height, dtype=np.float32)
            grid_w = np.arange(width, dtype=np.float32)
            grid = np.meshgrid(grid_w, grid_h)
            grid = np.stack(grid, axis=0).reshape(2, 1, height, width)
            pos_np = get_2d_sincos_pos_embed_from_grid(self.hidden_size_output, grid)
            pos = torch.from_numpy(pos_np).to(device=device, dtype=dtype)
            self._pos_cache[key] = pos
            return pos

    def forward(self, inputs: torch.Tensor, img_height: int = None, img_width: int = None, patch_size: int = None):
        if inputs.dim() != 4:
            raise ValueError("PixelTokenEmbedder expects inputs of shape [B,C,H,W]")
        assert img_height is not None and img_width is not None and patch_size is not None
        B, C, H, W = inputs.shape
        assert H == img_height and W == img_width
        assert (H % patch_size == 0) and (W % patch_size == 0)
        Hs, Ws = H // patch_size, W // patch_size
        P2 = patch_size * patch_size
        x = inputs.permute(0, 2, 3, 1).contiguous()
        x = self.proj(x)
        if self.use_pixel_abs_pos:
            pos_full = self._fetch_pixel_pos_image(H, W, inputs.device, inputs.dtype)
            pos_full = pos_full.view(H, W, self.hidden_size_output)
            x = x + pos_full.unsqueeze(0)
        x = x.view(B, Hs, patch_size, Ws, patch_size, self.hidden_size_output)
        x = x.permute(0, 1, 3, 2, 4, 5).contiguous()
        x = x.view(B * Hs * Ws, P2, self.hidden_size_output)
        return x


class PiTBlock(nn.Module):
    def __init__(
        self,
        pixel_hidden_size: int,
        patch_hidden_size: int,
        patch_size: int,
        num_heads: int,
        mlp_ratio: float = 4.0,
        attn_hidden_size: Optional[int] = None,
        attn_num_heads: Optional[int] = None,
        rope_fn=None,
    ):
        super().__init__()
        self.pixel_dim = int(pixel_hidden_size)
        self.context_dim = int(patch_hidden_size)
        self.patch_size = int(patch_size)
        self.attn_dim = int(attn_hidden_size) if attn_hidden_size is not None else self.context_dim
        self.num_heads = int(attn_num_heads) if attn_num_heads is not None else int(num_heads)
        assert (
            self.attn_dim % self.num_heads == 0
        ), "pixel attention hidden size must be divisible by pixel num_heads"
        p2 = self.patch_size * self.patch_size
        self.compress_to_attn = nn.Linear(p2 * self.pixel_dim, self.attn_dim, bias=True)
        self.expand_from_attn = nn.Linear(self.attn_dim, p2 * self.pixel_dim, bias=True)
        self.norm1 = RMSNorm(self.pixel_dim, eps=1e-6)
        self.attn = RotaryAttention(self.attn_dim, num_heads=self.num_heads, qkv_bias=False)
        self.norm2 = RMSNorm(self.pixel_dim, eps=1e-6)
        self.mlp = MLP(self.pixel_dim, mlp_ratio=mlp_ratio, drop=0.0)
        self.adaLN_modulation = nn.Sequential(nn.Linear(self.context_dim, 6 * self.pixel_dim * p2, bias=True))
        self._pos_cache = dict()
        self._rope_fn = rope_fn if rope_fn is not None else precompute_freqs_cis_2d

    def _fetch_pos(self, height: int, width: int, device):
        key = (height, width)
        if key in self._pos_cache:
            return self._pos_cache[key].to(device)
        pos = self._rope_fn(self.attn_dim // self.num_heads, height, width).to(device)
        self._pos_cache[key] = pos
        return pos

    def forward(self, x: torch.Tensor, s_cond: torch.Tensor, image_height: int, image_width: int, patch_size: int, mask=None) -> torch.Tensor:
        BL, P2, C = x.shape
        if C != self.pixel_dim:
            raise ValueError(f"PiTBlock expected pixel_dim={self.pixel_dim}, got {C}")
        assert (image_height % patch_size == 0) and (image_width % patch_size == 0)
        Hs, Ws = image_height // patch_size, image_width // patch_size
        L = Hs * Ws
        B = BL // L
        cond_params = self.adaLN_modulation(s_cond)
        cond_params = cond_params.view(BL, P2, 6 * self.pixel_dim)
        shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = torch.chunk(cond_params, 6, dim=-1)
        x_norm = apply_adaln(self.norm1(x), shift_msa, scale_msa)
        x_flat = x_norm.view(BL, P2 * self.pixel_dim)
        x_comp = self.compress_to_attn(x_flat).view(B, L, self.attn_dim)
        pos_comp = self._fetch_pos(Hs, Ws, x.device)
        attn_out = self.attn(x_comp, pos_comp, mask)
        attn_flat = self.expand_from_attn(attn_out.view(B * L, self.attn_dim))
        attn_exp = attn_flat.view(BL, P2, self.pixel_dim)
        x = x + gate_msa * attn_exp
        mlp_out = self.mlp(apply_adaln(self.norm2(x), shift_mlp, scale_mlp))
        x = x + gate_mlp * mlp_out
        return x


class PixDiT(nn.Module):
    def __init__(
        self,
        in_channels=4,
        num_groups=12,
        hidden_size=1152,
        pixel_hidden_size=64,
        patch_depth=18,
        pixel_depth=4,
        patch_size=2,
        num_classes=1000,
        use_pixel_abs_pos=True,
    ):
        super().__init__()
        self.in_channels = int(in_channels)
        self.out_channels = int(in_channels)
        self.hidden_size = int(hidden_size)
        self.num_groups = int(num_groups)
        self.patch_depth = int(patch_depth)
        self.pixel_depth = int(pixel_depth)
        self.patch_size = int(patch_size)
        self.pixel_hidden_size = int(pixel_hidden_size)
        self.num_classes = int(num_classes)
        self.use_pixel_abs_pos = bool(use_pixel_abs_pos)
        if self.pixel_depth <= 0:
            raise ValueError("PixDiT expects pixel_depth > 0 to preserve the dual-level pipeline")

        self.pixel_embedder = PixelTokenEmbedder(self.in_channels, self.pixel_hidden_size, use_pixel_abs_pos=self.use_pixel_abs_pos)
        self.s_embedder = PatchTokenEmbedder(self.in_channels * self.patch_size ** 2, self.hidden_size, bias=True)
        self.t_embedder = TimestepConditioner(self.hidden_size)
        self.y_embedder = ClassEmbedder(self.num_classes + 1, self.hidden_size)

        self.final_layer = FinalLayer(self.pixel_hidden_size, self.out_channels)
        self.patch_blocks = nn.ModuleList(
            [AugmentedDiTBlock(self.hidden_size, self.num_groups) for _ in range(self.patch_depth)]
        )
        self.pixel_blocks = nn.ModuleList(
            [
                PiTBlock(
                    self.pixel_hidden_size,
                    self.hidden_size,
                    patch_size=self.patch_size,
                    num_heads=self.num_groups,
                    mlp_ratio=4.0,
                )
                for _ in range(self.pixel_depth)
            ]
        )
        self.initialize_weights()
        self.precompute_pos = dict()

    def fetch_pos(self, height, width, device):
        if (height, width) in self.precompute_pos:
            return self.precompute_pos[(height, width)].to(device)
        else:
            pos = precompute_freqs_cis_2d(self.hidden_size // self.num_groups, height, width).to(device)
            self.precompute_pos[(height, width)] = pos
            return pos

    def initialize_weights(self):
        w = self.s_embedder.proj.weight.data
        nn.init.xavier_uniform_(w.view([w.shape[0], -1]))
        nn.init.constant_(self.s_embedder.proj.bias, 0)
        nn.init.normal_(self.y_embedder.embedding_table.weight, std=0.02)
        nn.init.normal_(self.t_embedder.mlp[0].weight, std=0.02)
        nn.init.normal_(self.t_embedder.mlp[2].weight, std=0.02)
        nn.init.zeros_(self.final_layer.linear.weight)
        nn.init.zeros_(self.final_layer.linear.bias)
        for block in self.patch_blocks:
            nn.init.zeros_(block.adaLN_modulation[0].weight)
            nn.init.zeros_(block.adaLN_modulation[0].bias)
        for block in self.pixel_blocks:
            nn.init.zeros_(block.adaLN_modulation[0].weight)
            nn.init.zeros_(block.adaLN_modulation[0].bias)

    def forward(self, x, t, y, s=None, mask=None):
        B, _, H, W = x.shape
        pos = self.fetch_pos(H // self.patch_size, W // self.patch_size, x.device)
        x_patches = torch.nn.functional.unfold(x, kernel_size=self.patch_size, stride=self.patch_size).transpose(1, 2)
        t_emb = self.t_embedder(t.view(-1)).view(B, -1, self.hidden_size)
        y_emb = self.y_embedder(y).view(B, 1, self.hidden_size)
        c = nn.functional.silu(t_emb + y_emb)
        if s is None:
            s = self.s_embedder(x_patches)
            for block in self.patch_blocks:
                s = block(s, c, pos, mask)
            s = nn.functional.silu(t_emb + s)
        batch_size, length, _ = s.shape
        s_cond = s.view(batch_size * length, self.hidden_size)
        x_pixels = self.pixel_embedder(x, img_height=H, img_width=W, patch_size=self.patch_size)
        for blk in self.pixel_blocks:
            x_pixels = blk(x_pixels, s_cond, H, W, self.patch_size, mask)
        x_pixels = self.final_layer(x_pixels)
        C_out = self.out_channels
        P2 = self.patch_size * self.patch_size
        x_pixels = x_pixels.view(B, length, P2, C_out).permute(0, 3, 2, 1).contiguous()
        x_pixels = x_pixels.view(B, C_out * P2, length)
        x_img = torch.nn.functional.fold(x_pixels, (H, W), kernel_size=self.patch_size, stride=self.patch_size)
        return x_img