File size: 6,047 Bytes
f1df74a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import logging

import torch
from einops import rearrange, repeat

from lvdm.models.utils_diffusion import timestep_embedding

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

mainlogger = logging.getLogger('mainlogger')



def TemporalTransformer_forward(self, x, context=None, is_imgbatch=False):
    b, c, t, h, w = x.shape
    x_in = x
    x = self.norm(x)
    x = rearrange(x, 'b c t h w -> (b h w) c t').contiguous()
    if not self.use_linear:
        x = self.proj_in(x)
    x = rearrange(x, 'bhw c t -> bhw t c').contiguous()
    if self.use_linear:
        x = self.proj_in(x)

    temp_mask = None
    if self.causal_attention:
        temp_mask = torch.tril(torch.ones([1, t, t]))
    if is_imgbatch:
        temp_mask = torch.eye(t).unsqueeze(0)
    if temp_mask is not None:
        mask = temp_mask.to(x.device)
        mask = repeat(mask, 'l i j -> (l bhw) i j', bhw=b*h*w)
    else:
        mask = None

    if self.only_self_att:
        ## note: if no context is given, cross-attention defaults to self-attention
        for i, block in enumerate(self.transformer_blocks):
            x = block(x, context=context, mask=mask)
        x = rearrange(x, '(b hw) t c -> b hw t c', b=b).contiguous()
    else:
        x = rearrange(x, '(b hw) t c -> b hw t c', b=b).contiguous()
        context = rearrange(context, '(b t) l con -> b t l con', t=t).contiguous()
        for i, block in enumerate(self.transformer_blocks):
            # calculate each batch one by one (since number in shape could not greater then 65,535 for some package)
            for j in range(b):
                unit_context = context[j][0:1]
                context_j = repeat(unit_context, 't l con -> (t r) l con', r=(h * w)).contiguous()
                ## note: causal mask will not applied in cross-attention case
                x[j] = block(x[j], context=context_j)
    
    if self.use_linear:
        x = self.proj_out(x)
        x = rearrange(x, 'b (h w) t c -> b c t h w', h=h, w=w).contiguous()
    if not self.use_linear:
        x = rearrange(x, 'b hw t c -> (b hw) c t').contiguous()
        x = self.proj_out(x)
        x = rearrange(x, '(b h w) c t -> b c t h w', b=b, h=h, w=w).contiguous()

    if self.use_image_dataset:
        x = 0.0 * x + x_in
    else:
        x = x + x_in
    return x

def selfattn_forward_unet(self, x, timesteps, context=None, y=None, features_adapter=None, is_imgbatch=False, T=None,  **kwargs):
        b,_,t,_,_ = x.shape
    
        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
        emb = self.time_embed(t_emb)
        if self.micro_condition and y is not None:
            micro_emb = timestep_embedding(y, self.model_channels, repeat_only=False)
            emb = emb + self.micro_embed(micro_emb)

        

        # pose_emb = pose_emb.reshape(-1, pose_emb.shape[-1])
        ## repeat t times for context [(b t) 77 768] & time embedding
        if not is_imgbatch:
            context = context.repeat_interleave(repeats=t, dim=0)

        if 'pose_emb' in kwargs:
            pose_emb = kwargs.pop('pose_emb')
            context = { 'context': context, 'pose_emb': pose_emb }

        emb = emb.repeat_interleave(repeats=t, dim=0)

        ## always in shape (b t) c h w, except for temporal layer
        x = rearrange(x, 'b c t h w -> (b t) c h w')
        if features_adapter is not None:
            features_adapter = [rearrange(feature, 'b c t h w -> (b t) c h w') for feature in features_adapter]

        h = x.type(self.dtype)
        adapter_idx = 0
        hs = []
        for id, module in enumerate(self.input_blocks):
            h = module(h, emb, context=context, batch_size=b,is_imgbatch=is_imgbatch)
            if id ==0 and self.addition_attention:
                h = self.init_attn(h, emb, context=context, batch_size=b,is_imgbatch=is_imgbatch)
            ## plug-in adapter features
            if ((id+1)%3 == 0) and features_adapter is not None:
                # if adapter_idx == 0 or adapter_idx == 1 or adapter_idx == 2:
                h = h + features_adapter[adapter_idx]
                adapter_idx += 1
            hs.append(h)
        if features_adapter is not None:
            assert len(features_adapter)==adapter_idx, 'Wrong features_adapter'

        h = self.middle_block(h, emb, context=context, batch_size=b, is_imgbatch=is_imgbatch)
        for module in self.output_blocks:
            h = torch.cat([h, hs.pop()], dim=1)
            h = module(h, emb, context=context, batch_size=b, is_imgbatch=is_imgbatch)
        h = h.type(x.dtype)
        y = self.out(h)
        
        # reshape back to (b c t h w)
        y = rearrange(y, '(b t) c h w -> b c t h w', b=b)
        return y

def spatial_forward_BasicTransformerBlock(self, x, context=None, mask=None):
    if isinstance(context, dict):
        context = context['context']
    x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None, mask=mask) + x
    x = self.attn2(self.norm2(x), context=context, mask=mask) + x
    x = self.ff(self.norm3(x)) + x
    return x

def temporal_selfattn_forward_BasicTransformerBlock(self, x, context=None, mask=None):
    if isinstance(context, dict) and 'pose_emb' in context:
        pose_emb = context['pose_emb'] # {channel_num: [B, video_length, pose_dim, pose_embedding_dim]}
        context = None
    else:
        pose_emb = None
        context = None

    x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None, mask=mask) + x

    # Add camera pose
    if pose_emb is not None:
        B, t, _, _ = pose_emb.shape # [B, video_length, pose_dim, pose_embedding_dim]
        hw = x.shape[0] // B
        pose_emb = pose_emb.reshape(B, t, -1)
        pose_emb = pose_emb.repeat_interleave(repeats=hw, dim=0)
        x = self.cc_projection(torch.cat([x, pose_emb], dim=-1))

    x = self.attn2(self.norm2(x), context=context, mask=mask) + x
    x = self.ff(self.norm3(x)) + x
    return x