File size: 5,011 Bytes
7f51798
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch.nn as nn
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 pdb import set_trace as st

from ldm.modules.attention import MemoryEfficientCrossAttention
from .dit_decoder import DiT2

class DiT3D(DiT2):
    def __init__(self, input_size=32, patch_size=2, in_channels=4, hidden_size=1152, depth=28, num_heads=16, mlp_ratio=4, class_dropout_prob=0.1, num_classes=1000, learn_sigma=True, mixing_logit_init=-3, mixed_prediction=True, context_dim=False, roll_out=False, plane_n=3, return_all_layers=False, in_plane_attention=True, vit_blk=...):
        super().__init__(input_size, patch_size, in_channels, hidden_size, depth, num_heads, mlp_ratio, class_dropout_prob, num_classes, learn_sigma, mixing_logit_init, mixed_prediction, context_dim, roll_out, plane_n, return_all_layers, in_plane_attention, vit_blk)
        # follow point infinity, add "write" CA block per 6 blocks

        # 25/4/2024, cascade a "read&write" block after the DiT base model.
        self.read_ca = MemoryEfficientCrossAttention(hidden_size, context_dim)
        self.point_infinity_blocks = nn.ModuleList([
            vit_blk(hidden_size, num_heads, mlp_ratio=mlp_ratio)
            for _ in range(2)
        ])
    
    def initialize_weights(self):
        super().initialize_weights()

        # Zero-out adaLN modulation layers in DiT blocks:
        # ! no final layer anymore
        for block in self.point_infinity_blocks:
            nn.init.constant_(block.adaLN_modulation[-1].weight, 0)
            nn.init.constant_(block.adaLN_modulation[-1].bias, 0)
                                                      
        
    
    def forward(self, c, *args, **kwargs):
        x_base = super().forward(c, *args, **kwargs) # base latent
        # add read&write block



#################################################################################
#                                   DiT3D Configs                                  #
#################################################################################


def DiT3DXL_2(**kwargs):
    return DiT3D(depth=28,
                hidden_size=1152,
                patch_size=2,
                num_heads=16,
                **kwargs)


def DiT3DXL_2_half(**kwargs):
    return DiT3D(depth=28 // 2,
                hidden_size=1152,
                patch_size=2,
                num_heads=16,
                **kwargs)


def DiT3DXL_4(**kwargs):
    return DiT3D(depth=28,
                hidden_size=1152,
                patch_size=4,
                num_heads=16,
                **kwargs)


def DiT3DXL_8(**kwargs):
    return DiT3D(depth=28,
                hidden_size=1152,
                patch_size=8,
                num_heads=16,
                **kwargs)


def DiT3DL_2(**kwargs):
    return DiT3D(depth=24,
                hidden_size=1024,
                patch_size=2,
                num_heads=16,
                **kwargs)


def DiT3DL_2_half(**kwargs):
    return DiT3D(depth=24 // 2,
                hidden_size=1024,
                patch_size=2,
                num_heads=16,
                **kwargs)


def DiT3DL_4(**kwargs):
    return DiT3D(depth=24,
                hidden_size=1024,
                patch_size=4,
                num_heads=16,
                **kwargs)


def DiT3DL_8(**kwargs):
    return DiT3D(depth=24,
                hidden_size=1024,
                patch_size=8,
                num_heads=16,
                **kwargs)


def DiT3DB_2(**kwargs):
    return DiT3D(depth=12,
                hidden_size=768,
                patch_size=2,
                num_heads=12,
                **kwargs)


def DiT3DB_4(**kwargs):
    return DiT3D(depth=12,
                hidden_size=768,
                patch_size=4,
                num_heads=12,
                **kwargs)


def DiT3DB_8(**kwargs):
    return DiT3D(depth=12,
                hidden_size=768,
                patch_size=8,
                num_heads=12,
                **kwargs)


def DiT3DB_16(**kwargs):  # ours cfg
    return DiT3D(depth=12,
                hidden_size=768,
                patch_size=16,
                num_heads=12,
                **kwargs)


def DiT3DS_2(**kwargs):
    return DiT3D(depth=12, hidden_size=384, patch_size=2, num_heads=6, **kwargs)


def DiT3DS_4(**kwargs):
    return DiT3D(depth=12, hidden_size=384, patch_size=4, num_heads=6, **kwargs)


def DiT3DS_8(**kwargs):
    return DiT3D(depth=12, hidden_size=384, patch_size=8, num_heads=6, **kwargs)


DiT3Dmodels = {
    'DiT3D-XL/2': DiT3DXL_2,
    'DiT3D-XL/2/half': DiT3DXL_2_half,
    'DiT3D-XL/4': DiT3DXL_4,
    'DiT3D-XL/8': DiT3DXL_8,
    'DiT3D-L/2': DiT3DL_2,
    'DiT3D-L/2/half': DiT3DL_2_half,
    'DiT3D-L/4': DiT3DL_4,
    'DiT3D-L/8': DiT3DL_8,
    'DiT3D-B/2': DiT3DB_2,
    'DiT3D-B/4': DiT3DB_4,
    'DiT3D-B/8': DiT3DB_8,
    'DiT3D-B/16': DiT3DB_16,
    'DiT3D-S/2': DiT3DS_2,
    'DiT3D-S/4': DiT3DS_4,
    'DiT3D-S/8': DiT3DS_8,
}