Datasets:

ArXiv:
File size: 19,739 Bytes
7b4c6f4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
import math
from typing import Optional, List, Union, Tuple

import torch
import torch.nn as nn
import torch.nn.functional as F

from models.helpers import DropPath, drop_path

from utils.model_args import ModelArgs
from transformers import AutoImageProcessor, AutoModel

# this file only provides the 3 blocks used in VAR transformer
__all__ = ['FFN', 'AdaLNSelfAttn', 'AdaLNBeforeHead']


# automatically import fused operators
dropout_add_layer_norm = fused_mlp_func = memory_efficient_attention = flash_attn_func = None
try:
    from flash_attn.ops.layer_norm import dropout_add_layer_norm
    from flash_attn.ops.fused_dense import fused_mlp_func
except ImportError: pass
# automatically import faster attention implementations
try: from xformers.ops import memory_efficient_attention
except ImportError: pass
try: from flash_attn import flash_attn_func              # qkv: BLHc, ret: BLHcq
except ImportError: pass
try: from torch.nn.functional import scaled_dot_product_attention as slow_attn    # q, k, v: BHLc
except ImportError:
    def slow_attn(query, key, value, scale: float, attn_mask=None, dropout_p=0.0):
        attn = query.mul(scale) @ key.transpose(-2, -1) # BHLc @ BHcL => BHLL
        if attn_mask is not None: attn.add_(attn_mask)
        return (F.dropout(attn.softmax(dim=-1), p=dropout_p, inplace=True) if dropout_p > 0 else attn.softmax(dim=-1)) @ value




class ConditionEmbedder(nn.Module):
    """
    Embeds Condition into vector representations. Also handles label dropout for classifier-free guidance.
    """
    def __init__(self, in_channels, hidden_size, uncond_prob, token_num=120, vocab_size=16384):
        super().__init__()
        self.cap_proj = MLP(in_features=hidden_size, hidden_features=hidden_size, out_features=hidden_size)
        self.register_buffer("uncond_embedding", torch.zeros(token_num, hidden_size) / hidden_size ** 0.5)
        self.uncond_prob = uncond_prob

    def token_drop(self, caption, force_drop_ids=None, drop_ids=None):
        """
        Drops labels to enable classifier-free guidance.
        """
        if force_drop_ids is None:
            if drop_ids is None:
                drop_ids = torch.rand(caption.shape[0], device=caption.device) < self.uncond_prob
        else:
            drop_ids = force_drop_ids == 1
        if self.uncond_embedding.shape[0] < caption.shape[1]:
            # 动态扩展
            repeat_factor = int(caption.shape[1] / self.uncond_embedding.shape[0]) + 1
            extended = self.uncond_embedding.repeat(repeat_factor, 1)[:caption.shape[1]]
        else:
            extended = self.uncond_embedding[:caption.shape[1]]

        caption = torch.where(drop_ids[:, None, None], extended, caption)

        # caption = torch.where(drop_ids[:, None, None], self.uncond_embedding[:caption.shape[1]], caption)
        return caption

    def forward(self, caption, train, force_drop_ids=None, drop_ids=None):
        use_dropout = self.uncond_prob > 0
        if (train and use_dropout) or (force_drop_ids is not None):
            caption = self.token_drop(caption, force_drop_ids, drop_ids)
        embeddings = self.cap_proj(caption)
        return embeddings

class MLP(nn.Module):
    def __init__(self, in_features, hidden_features, out_features):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features, bias=False)
        self.act = nn.GELU(approximate='tanh')
        self.fc2 = nn.Linear(hidden_features, out_features, bias=False)
        
        nn.init.zeros_(self.fc1.weight)
        nn.init.zeros_(self.fc2.weight)

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.fc2(x)
        return x


class Dinov2_Adapter(nn.Module):
    def __init__(self, input_dim=1, output_dim=768, attention=False, pool=False, nheads=8, dropout=0.1, adapter_size='small', condition_type='seg'):
        super(Dinov2_Adapter, self).__init__()
        # print(f"Choose adapter size: {adapter_size}")
        # print(f"condition type: {condition_type}")

        from transformers import logging
        logging.set_verbosity_error()



        self.model = AutoModel.from_pretrained(f'./dinov2_small',local_files_only=True, use_safetensors=False)
        self.condition_type = condition_type
    
    def to_patch14(self, input):
        H, W = input.shape[2:]
        new_H = (H // 16) * 14
        new_W = (W // 16) * 14
        if self.condition_type in ['canny', 'seg']:
            output = torch.nn.functional.interpolate(input, size=(new_H, new_W), mode='nearest')#, align_corners=True)  canny, seg
        else:
            output = torch.nn.functional.interpolate(input, size=(new_H, new_W), mode='bicubic', align_corners=True) # depth, lineart, hed
        return output
        
    def forward(self, x):
        x = self.to_patch14(x)
        x = self.model(x)
        return x.last_hidden_state[:, 1:]


#################################################################################
#                         Cross-Attention    Injection                          #
#################################################################################
class CrossAttentionInjection(nn.Module):
    def __init__(self, embed_dim, num_heads):
        super().__init__()
        self.query_proj = nn.Linear(embed_dim, embed_dim)
        self.key_proj = nn.Linear(embed_dim, embed_dim)
        self.value_proj = nn.Linear(embed_dim, embed_dim)
        self.out_proj = nn.Linear(embed_dim, embed_dim)
        self.num_heads = num_heads
        self.scale = (embed_dim // num_heads) ** -0.5

    def forward(self, x, cond_feat):
        """
        x: [B, L, C],主特征序列
        cond_feat: [B, L_cond, C],来自分割图的条件token序列
        """
        B, L, C = x.shape
        H = self.num_heads
        Q = self.query_proj(x).reshape(B, L, H, C // H).transpose(1, 2)  # [B, H, L, d]
        K = self.key_proj(cond_feat).reshape(B, -1, H, C // H).transpose(1, 2)  # [B, H, Lc, d]
        V = self.value_proj(cond_feat).reshape(B, -1, H, C // H).transpose(1, 2)  # [B, H, Lc, d]

        attn = (Q @ K.transpose(-2, -1)) * self.scale  # [B, H, L, Lc]
        attn = attn.softmax(dim=-1)
        out = (attn @ V).transpose(1, 2).reshape(B, L, C)  # [B, L, C]
        return self.out_proj(out)



class FFN(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, drop=0., fused_if_available=True):
        super().__init__()
        self.fused_mlp_func = fused_mlp_func if fused_if_available else None
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.act = nn.GELU(approximate='tanh')
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop, inplace=True) if drop > 0 else nn.Identity()
    
    def forward(self, x):
        if self.fused_mlp_func is not None:
            return self.drop(self.fused_mlp_func(
                x=x, weight1=self.fc1.weight, weight2=self.fc2.weight, bias1=self.fc1.bias, bias2=self.fc2.bias,
                activation='gelu_approx', save_pre_act=self.training, return_residual=False, checkpoint_lvl=0,
                heuristic=0, process_group=None,
            ))
        else:
            return self.drop(self.fc2( self.act(self.fc1(x)) ))
    
    def extra_repr(self) -> str:
        return f'fused_mlp_func={self.fused_mlp_func is not None}'

 
class SelfAttention(nn.Module):
    def __init__(
        self, block_idx, embed_dim=768, num_heads=12,
        attn_drop=0., proj_drop=0., attn_l2_norm=False, flash_if_available=False,
    ):
        super().__init__()
        assert embed_dim % num_heads == 0
        self.block_idx, self.num_heads, self.head_dim = block_idx, num_heads, embed_dim // num_heads  # =64
        self.attn_l2_norm = attn_l2_norm
        if self.attn_l2_norm:
            self.scale = 1
            self.scale_mul_1H11 = nn.Parameter(torch.full(size=(1, self.num_heads, 1, 1), fill_value=4.0).log(), requires_grad=True)
            self.max_scale_mul = torch.log(torch.tensor(100)).item()
        else:
            self.scale = 0.25 / math.sqrt(self.head_dim)
        
        self.mat_qkv = nn.Linear(embed_dim, embed_dim * 3, bias=False)
        self.q_bias, self.v_bias = nn.Parameter(torch.zeros(embed_dim)), nn.Parameter(torch.zeros(embed_dim))
        self.register_buffer('zero_k_bias', torch.zeros(embed_dim))
        
        self.proj = nn.Linear(embed_dim, embed_dim)
        self.proj_drop = nn.Dropout(proj_drop, inplace=True) if proj_drop > 0 else nn.Identity()
        self.attn_drop: float = attn_drop
        self.using_flash = flash_if_available and flash_attn_func is not None
        # self.using_xform = flash_if_available and memory_efficient_attention is not None
        self.using_xform = False
        # only used during inference
        self.caching, self.cached_k, self.cached_v = False, None, None
    
    def kv_caching(self, enable: bool): self.caching, self.cached_k, self.cached_v = enable, None, None
    
    # NOTE: attn_bias is None during inference because kv cache is enabled
    def forward(self, x, attn_bias):
        B, L, C = x.shape
        
        qkv = F.linear(input=x, weight=self.mat_qkv.weight, bias=torch.cat((self.q_bias, self.zero_k_bias, self.v_bias))).view(B, L, 3, self.num_heads, self.head_dim)
        main_type = qkv.dtype
        # qkv: BL3Hc
        
        using_flash = self.using_flash and attn_bias is None and qkv.dtype != torch.float32
        if using_flash or self.using_xform: q, k, v = qkv.unbind(dim=2); dim_cat = 1   # q or k or v: BLHc
        else: q, k, v = qkv.permute(2, 0, 3, 1, 4).unbind(dim=0); dim_cat = 2               # q or k or v: BHLc
        
        if self.attn_l2_norm:
            scale_mul = self.scale_mul_1H11.clamp_max(self.max_scale_mul).exp()
            if using_flash or self.using_xform: scale_mul = scale_mul.transpose(1, 2)  # 1H11 to 11H1
            q = F.normalize(q, dim=-1).mul(scale_mul)
            k = F.normalize(k, dim=-1)
        
        if self.caching:
            if self.cached_k is None: self.cached_k = k; self.cached_v = v
            else: k = self.cached_k = torch.cat((self.cached_k, k), dim=dim_cat); v = self.cached_v = torch.cat((self.cached_v, v), dim=dim_cat)
        
        dropout_p = self.attn_drop if self.training else 0.0
        if using_flash:
            oup = flash_attn_func(q.to(dtype=main_type), k.to(dtype=main_type), v.to(dtype=main_type), dropout_p=dropout_p, softmax_scale=self.scale).view(B, L, C)
        elif self.using_xform:
            oup = memory_efficient_attention(q.to(dtype=main_type), k.to(dtype=main_type), v.to(dtype=main_type), attn_bias=None if attn_bias is None else attn_bias.to(dtype=main_type).expand(B, self.num_heads, -1, -1), p=dropout_p, scale=self.scale).view(B, L, C)
        else:
            oup = slow_attn(query=q, key=k, value=v, scale=self.scale, attn_mask=attn_bias, dropout_p=dropout_p).transpose(1, 2).reshape(B, L, C)
        
        return self.proj_drop(self.proj(oup))
        # attn = (q @ k.transpose(-2, -1)).add_(attn_bias + self.local_rpb())  # BHLc @ BHcL => BHLL
        # attn = self.attn_drop(attn.softmax(dim=-1))
        # oup = (attn @ v).transpose_(1, 2).reshape(B, L, -1)     # BHLL @ BHLc = BHLc => BLHc => BLC
    
    def extra_repr(self) -> str:
        return f'using_flash={self.using_flash}, using_xform={self.using_xform}, attn_l2_norm={self.attn_l2_norm}'

config = ModelArgs()
class AdaLNSelfAttn(nn.Module):
    def __init__(
        self, block_idx, last_drop_p, embed_dim, cond_dim, shared_aln: bool, norm_layer,
        num_heads, mlp_ratio=4., drop=0., attn_drop=0., drop_path=0., attn_l2_norm=False,
        flash_if_available=False, fused_if_available=True,depth=16,
    ):
        super(AdaLNSelfAttn, self).__init__()
        self.block_idx, self.last_drop_p, self.C = block_idx, last_drop_p, embed_dim
        self.C, self.D = embed_dim, cond_dim
        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        self.attn = SelfAttention(block_idx=block_idx, embed_dim=embed_dim, num_heads=num_heads, attn_drop=attn_drop, proj_drop=drop, attn_l2_norm=attn_l2_norm, flash_if_available=flash_if_available)
        self.ffn = FFN(in_features=embed_dim, hidden_features=round(embed_dim * mlp_ratio), drop=drop, fused_if_available=fused_if_available)
        
        self.ln_wo_grad = norm_layer(embed_dim, elementwise_affine=False)
        self.shared_aln = shared_aln
        if self.shared_aln:
            self.ada_gss = nn.Parameter(torch.randn(1, 1, 6, embed_dim) / embed_dim**0.5)
        else:
            lin = nn.Linear(cond_dim, 6*embed_dim)
            self.ada_lin = nn.Sequential(nn.SiLU(inplace=False), lin)
        
        self.fused_add_norm_fn = None

        self.adapter = Dinov2_Adapter(adapter_size=config.adapter_size, condition_type=config.condition_type)
        #  冻结 Dinov2 主干权重
        for p in self.adapter.model.parameters():
            p.requires_grad = False
        
        # self.adapter = EVA_Adapter()
        if config.adapter_size == "small":
            self.adapter_mlp = MLP(384, config.dim, config.dim)
        elif config.adapter_size == 'base':
            self.adapter_mlp = MLP(768, config.dim, config.dim)

        self.condition_embeddings = nn.Embedding(config.vocab_size, config.dim)
        self.condition_mlp = ConditionEmbedder(config.block_size, config.dim, config.class_dropout_prob, config.block_size, config.vocab_size)
        # conditon 注入层
        self.condition_layers = torch.nn.ModuleList()
        for layer_id in range(3):
            self.condition_layers.append(MLP(config.dim,config.dim,config.dim))
        
        self.layer_internal = depth=16 // 2
        self.control_strength = 1  

        #CrossAttention注入方式
        self.cross_attn_inject = CrossAttentionInjection(embed_dim=config.dim, num_heads=num_heads)


    # NOTE: attn_bias is None during inference because kv cache is enabled
    def forward(self, x, cond_BD, condition, attn_bias, current_step: int, total_steps:int):   # C: embed_dim, D: cond_dim
        if self.shared_aln:
            gamma1, gamma2, scale1, scale2, shift1, shift2 = (self.ada_gss + cond_BD).unbind(2) # 116C + B16C =unbind(2)=> 6 B1C
        else:
            gamma1, gamma2, scale1, scale2, shift1, shift2 = self.ada_lin(cond_BD).view(-1, 1, 6, self.C).unbind(2)
    
        # --------- 注入策略一:条件注入机制(例如每隔 N 层注入) ---------
        # if condition is not None:
        #     condition_embeddings = self.adapter(condition)
        #     condition_embeddings = self.adapter_mlp(condition_embeddings)
        #     self.condition_token = self.condition_mlp(condition_embeddings,train=self.training)

        # # self.block_idx 是当前block编号
        # if self.block_idx % self.layer_internal == 0:
        #     cond_feat = self.condition_layers[self.block_idx // self.layer_internal](self.condition_token)  # [B, 1, C]
        #     # cond_feat: [B, Lc, C] → [B, Lx, C]
        #     cond_feat = cond_feat.mean(dim=1, keepdim=True).expand(-1, x.shape[1], -1)
        #     x = x + self.control_strength * cond_feat

        # # --------- 注入策略二:在某些层激活 cross-attention 注入(例如每隔 N 层注入) ---------
        # if condition is not None:
        #     condition_embeddings = self.adapter(condition)
        #     condition_embeddings = self.adapter_mlp(condition_embeddings)
        #     self.condition_token = self.condition_mlp(condition_embeddings,train=self.training)
            
        # if self.block_idx % self.layer_internal == 0:
        #     cond_feat = self.condition_layers[self.block_idx // self.layer_internal](self.condition_token)
        #     cond_feat = cond_feat.mean(dim=1, keepdim=True).expand(-1, x.shape[1], -1)
        #     x = x + self.control_strength * cond_feat

        #     # cross-attention: x attends to condition token
        #     # query: x, key/value: condition_token
        #     cross_attn_out = self.cross_attn_inject(query=x, key=self.condition_token, value=self.condition_token)
        #     x = x + self.control_strength * cross_attn_out


        # --------- 注入策略三:注入强度控制机制 ---------
        # 使用训练步调度:
        if condition is not None:
            condition_embeddings = self.adapter(condition)
            condition_embeddings = self.adapter_mlp(condition_embeddings)
            self.condition_token = self.condition_mlp(condition_embeddings,train=self.training)
            
            cond_feat = self.condition_layers[self.block_idx // self.layer_internal](self.condition_token)
            cond_feat = cond_feat.mean(dim=1, keepdim=True).expand(-1, x.shape[1], -1)
           
            # cross_attn_out = self.cross_attn_inject(x, key=self.condition_token, value=self.condition_token)
            cross_attn_out = self.cross_attn_inject(x, self.condition_token)

        if current_step is not None:
            progress = min(current_step / total_steps, 1.0)
            alpha = 0.5 * (1 + math.cos(math.pi * progress))
        else:
            alpha = 1.0
            
        x = x + alpha * cross_attn_out

        # --------- 注入策略四:区域掩码控制机制 ---------
        # 在生成的早期阶段,边缘信息可能会有更大的权重来帮助模型生成图像的轮廓
        # 在生成的后期,分割图的控制可能更重要,用于确保图像区域的语义一致性 
        # progress = current_step / total_steps

        # alpha_edge = 1.0 - progress     # 边缘控制随时间递减
        # alpha_seg  = progress           # 分割控制随时间增强

        # edge_feat = self.edge_adapter(edge_map) # 这两个部分需要处理
        # seg_feat  = self.seg_adapter(seg_map)   # 需要边缘提取图,以及处理模型:ViT?????

        # # 将两个特征注入模型
        # x = x + alpha_edge * self.cross_attn_edge(x, edge_feat)
        # x = x + alpha_seg  * self.cross_attn_seg(x, seg_feat)


        # --------- 注入策略五:阶段性控制机制 ---------
        
        # 根据训练轮次逐步增强/弱化注入作用:warm-up、余弦下降、分阶段注入策略

        # if self.training:
        #     if current_epoch < 10:
        #         alpha = 0.1
        #     elif current_epoch < 30:
        #         alpha = 0.5
        #     else:
        #         alpha = 1.0
        # x = x + alpha * cross_attn_out


        # --------- 注意力 + FFN  ---------
        x = x + self.drop_path(self.attn( self.ln_wo_grad(x).mul(scale1.add(1)).add_(shift1), attn_bias=attn_bias ).mul_(gamma1))
        x = x + self.drop_path(self.ffn( self.ln_wo_grad(x).mul(scale2.add(1)).add_(shift2) ).mul(gamma2)) # this mul(gamma2) cannot be in-placed when FusedMLP is used
        return x


    def extra_repr(self) -> str:
        return f'shared_aln={self.shared_aln}'


class AdaLNBeforeHead(nn.Module):
    def __init__(self, C, D, norm_layer):   # C: embed_dim, D: cond_dim
        super().__init__()
        self.C, self.D = C, D
        self.ln_wo_grad = norm_layer(C, elementwise_affine=False)
        self.ada_lin = nn.Sequential(nn.SiLU(inplace=False), nn.Linear(D, 2*C))
    
    def forward(self, x_BLC: torch.Tensor, cond_BD: torch.Tensor):
        scale, shift = self.ada_lin(cond_BD).view(-1, 1, 2, self.C).unbind(2)
        return self.ln_wo_grad(x_BLC).mul(scale.add(1)).add_(shift)