File size: 15,969 Bytes
265ae36
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
# Copyright (c) Meta Platforms, Inc. and affiliates.
#
# This source code is licensed under the Apache License, Version 2.0
# found in the LICENSE file in the root directory of this source tree.

# References:
#   https://github.com/facebookresearch/dino/blob/main/vision_transformer.py
#   https://github.com/rwightman/pytorch-image-models/tree/master/timm/models/vision_transformer.py

from functools import partial
import math
import logging
from typing import Sequence, Tuple, Union, Callable

import torch
import torch.nn as nn
import torch.utils.checkpoint
from torch.nn.init import trunc_normal_

from src.dinov2.layers import Mlp, PatchEmbed, SwiGLUFFNFused, MemEffAttention, NestedTensorBlockVis as Block


logger = logging.getLogger("dinov2")


def named_apply(fn: Callable, module: nn.Module, name="", depth_first=True, include_root=False) -> nn.Module:
    if not depth_first and include_root:
        fn(module=module, name=name)
    for child_name, child_module in module.named_children():
        child_name = ".".join((name, child_name)) if name else child_name
        named_apply(fn=fn, module=child_module, name=child_name, depth_first=depth_first, include_root=True)
    if depth_first and include_root:
        fn(module=module, name=name)
    return module


class BlockChunk(nn.ModuleList):
    def forward(self, x):
        for b in self:
            x = b(x)
        return x


class DinoVisionTransformer(nn.Module):
    def __init__(
        self,
        img_size=518,
        patch_size=16,
        in_chans=3,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4.0,
        qkv_bias=True,
        ffn_bias=True,
        proj_bias=True,
        drop_path_rate=0.0,
        drop_path_uniform=False,
        init_values=None,  # for layerscale: None or 0 => no layerscale
        embed_layer=PatchEmbed,
        act_layer=nn.GELU,
        block_fn=Block,
        ffn_layer="mlp",
        block_chunks=1,
        num_register_tokens=0,
        interpolate_antialias=False,
        interpolate_offset=0.1,
    ):
        """
        Args:
            img_size (int, tuple): input image size
            patch_size (int, tuple): patch size
            in_chans (int): number of input channels
            embed_dim (int): embedding dimension
            depth (int): depth of transformer
            num_heads (int): number of attention heads
            mlp_ratio (int): ratio of mlp hidden dim to embedding dim
            qkv_bias (bool): enable bias for qkv if True
            proj_bias (bool): enable bias for proj in attn if True
            ffn_bias (bool): enable bias for ffn if True
            drop_path_rate (float): stochastic depth rate
            drop_path_uniform (bool): apply uniform drop rate across blocks
            weight_init (str): weight init scheme
            init_values (float): layer-scale init values
            embed_layer (nn.Module): patch embedding layer
            act_layer (nn.Module): MLP activation layer
            block_fn (nn.Module): transformer block class
            ffn_layer (str): "mlp", "swiglu", "swiglufused" or "identity"
            block_chunks: (int) split block sequence into block_chunks units for FSDP wrap
            num_register_tokens: (int) number of extra cls tokens (so-called "registers")
            interpolate_antialias: (str) flag to apply anti-aliasing when interpolating positional embeddings
            interpolate_offset: (float) work-around offset to apply when interpolating positional embeddings
        """
        super().__init__()
        norm_layer = partial(nn.LayerNorm, eps=1e-6)

        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
        self.num_tokens = 1
        self.n_blocks = depth
        self.num_heads = num_heads
        self.patch_size = patch_size
        self.num_register_tokens = num_register_tokens
        self.interpolate_antialias = interpolate_antialias
        self.interpolate_offset = interpolate_offset

        self.patch_embed = embed_layer(img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim)
        num_patches = self.patch_embed.num_patches

        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + self.num_tokens, embed_dim))
        assert num_register_tokens >= 0
        self.register_tokens = (
            nn.Parameter(torch.zeros(1, num_register_tokens, embed_dim)) if num_register_tokens else None
        )

        if drop_path_uniform is True:
            dpr = [drop_path_rate] * depth
        else:
            dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule

        if ffn_layer == "mlp":
            logger.info("using MLP layer as FFN")
            ffn_layer = Mlp
        elif ffn_layer == "swiglufused" or ffn_layer == "swiglu":
            logger.info("using SwiGLU layer as FFN")
            ffn_layer = SwiGLUFFNFused
        elif ffn_layer == "identity":
            logger.info("using Identity layer as FFN")

            def f(*args, **kwargs):
                return nn.Identity()

            ffn_layer = f
        else:
            raise NotImplementedError

        blocks_list = [
            block_fn(
                dim=embed_dim,
                num_heads=num_heads,
                mlp_ratio=mlp_ratio,
                qkv_bias=qkv_bias,
                proj_bias=proj_bias,
                ffn_bias=ffn_bias,
                drop_path=dpr[i],
                norm_layer=norm_layer,
                act_layer=act_layer,
                ffn_layer=ffn_layer,
                init_values=init_values,
            )
            for i in range(depth)
        ]
        if block_chunks > 0:
            self.chunked_blocks = True
            chunked_blocks = []
            chunksize = depth // block_chunks
            for i in range(0, depth, chunksize):
                # this is to keep the block index consistent if we chunk the block list
                chunked_blocks.append([nn.Identity()] * i + blocks_list[i : i + chunksize])
            self.blocks = nn.ModuleList([BlockChunk(p) for p in chunked_blocks])
        else:
            self.chunked_blocks = False
            self.blocks = nn.ModuleList(blocks_list)

        self.norm = norm_layer(embed_dim)
        self.head = nn.Identity()

        self.mask_token = nn.Parameter(torch.zeros(1, embed_dim))

        self.init_weights()

    def get_last_self_attention(self, x, masks=None):
        if isinstance(x, list):
            return self.forward_features_list(x, masks)
            
        x = self.prepare_tokens_with_masks(x, masks)
        
        # Run through model, at the last block just return the attention.
        for i, blk in enumerate(self.blocks):
            if i < len(self.blocks) - 1:
                x = blk(x)
            else: 
                return blk(x, return_attention=True)


    def init_weights(self):
        trunc_normal_(self.pos_embed, std=0.02)
        nn.init.normal_(self.cls_token, std=1e-6)
        if self.register_tokens is not None:
            nn.init.normal_(self.register_tokens, std=1e-6)
        named_apply(init_weights_vit_timm, self)

    def interpolate_pos_encoding(self, x, w, h):
        previous_dtype = x.dtype
        npatch = x.shape[1] - 1
        N = self.pos_embed.shape[1] - 1
        if npatch == N and w == h:
            return self.pos_embed
        pos_embed = self.pos_embed.float()
        class_pos_embed = pos_embed[:, 0]
        patch_pos_embed = pos_embed[:, 1:]
        dim = x.shape[-1]
        w0 = w // self.patch_size
        h0 = h // self.patch_size
        M = int(math.sqrt(N))  # Recover the number of patches in each dimension
        assert N == M * M
        kwargs = {}
        if self.interpolate_offset:
            # Historical kludge: add a small number to avoid floating point error in the interpolation, see https://github.com/facebookresearch/dino/issues/8
            # Note: still needed for backward-compatibility, the underlying operators are using both output size and scale factors
            sx = float(w0 + self.interpolate_offset) / M
            sy = float(h0 + self.interpolate_offset) / M
            kwargs["scale_factor"] = (sx, sy)
        else:
            # Simply specify an output size instead of a scale factor
            kwargs["size"] = (w0, h0)
        patch_pos_embed = nn.functional.interpolate(
            patch_pos_embed.reshape(1, M, M, dim).permute(0, 3, 1, 2),
            mode="bicubic",
            antialias=self.interpolate_antialias,
            **kwargs,
        )
        assert (w0, h0) == patch_pos_embed.shape[-2:]
        patch_pos_embed = patch_pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)
        return torch.cat((class_pos_embed.unsqueeze(0), patch_pos_embed), dim=1).to(previous_dtype)

    def prepare_tokens_with_masks(self, x, masks=None, prompt=None):
        # print("prompt", prompt)
        B, nc, w, h = x.shape
        x = self.patch_embed(x)
        
        if masks is not None:
            x = torch.where(masks.unsqueeze(-1), self.mask_token.to(x.dtype).unsqueeze(0), x)

        x = torch.cat((self.cls_token.expand(x.shape[0], -1, -1), x), dim=1)
        
        x = x + self.interpolate_pos_encoding(x, w, h)
        
        # if prompt is not None:
        #     x = torch.cat([x, prompt], dim=1)

        if prompt is not None:
            x = torch.cat(
                (
                    x[:, :1],
                    prompt,
                    x[:, 1:]
                ),
                dim=1
            )

        if self.register_tokens is not None:
            x = torch.cat(
                (
                    x[:, :1],
                    self.register_tokens.expand(x.shape[0], -1, -1),
                    x[:, 1:],
                ),
                dim=1,
            )

        return x

    def forward_features_list(self, x_list, masks_list, prompt_list):
        x = [self.prepare_tokens_with_masks(x, masks, prompt) for x, masks, prompt in zip(x_list, masks_list, prompt_list)]
        for blk in self.blocks:
            x = blk(x)

        all_x = x
        output = []
        for x, masks in zip(all_x, masks_list):
            x_norm = self.norm(x)
            output.append(
                {
                    "x_norm_clstoken": x_norm[:, 0],
                    "x_norm_regtokens": x_norm[:, 1 : self.num_register_tokens + 1],
                    "x_norm_patchtokens": x_norm[:, self.num_register_tokens + 1 :],
                    "x_prenorm": x,
                    "masks": masks,
                }
            )
        return output

    def forward_features(self, x, masks=None, prompt=None):
        # print("forward_features prompt: ", prompt)
        if isinstance(x, list):
            return self.forward_features_list(x, masks, prompt)

        x = self.prepare_tokens_with_masks(x, masks, prompt)

        for blk in self.blocks:
            x = blk(x)

        x_norm = self.norm(x)
        return {
            "x_norm_clstoken": x_norm[:, 0],
            "x_norm_regtokens": x_norm[:, 1 : self.num_register_tokens + 1],
            "x_norm_patchtokens": x_norm[:, self.num_register_tokens + 1 :],
            "x_prenorm": x,
            "masks": masks,
        }

    def _get_intermediate_layers_not_chunked(self, x, n=1):
        x = self.prepare_tokens_with_masks(x)
        # If n is an int, take the n last blocks. If it's a list, take them
        output, total_block_len = [], len(self.blocks)
        blocks_to_take = range(total_block_len - n, total_block_len) if isinstance(n, int) else n
        for i, blk in enumerate(self.blocks):
            x = blk(x)
            if i in blocks_to_take:
                output.append(x)
        assert len(output) == len(blocks_to_take), f"only {len(output)} / {len(blocks_to_take)} blocks found"
        return output

    def _get_intermediate_layers_chunked(self, x, n=1):
        x = self.prepare_tokens_with_masks(x)
        output, i, total_block_len = [], 0, len(self.blocks[-1])
        # If n is an int, take the n last blocks. If it's a list, take them
        blocks_to_take = range(total_block_len - n, total_block_len) if isinstance(n, int) else n
        for block_chunk in self.blocks:
            for blk in block_chunk[i:]:  # Passing the nn.Identity()
                x = blk(x)
                if i in blocks_to_take:
                    output.append(x)
                i += 1
        assert len(output) == len(blocks_to_take), f"only {len(output)} / {len(blocks_to_take)} blocks found"
        return output

    def get_intermediate_layers(
        self,
        x: torch.Tensor,
        n: Union[int, Sequence] = 1,  # Layers or n last layers to take
        reshape: bool = False,
        return_class_token: bool = False,
        norm=True,
    ) -> Tuple[Union[torch.Tensor, Tuple[torch.Tensor]]]:
        if self.chunked_blocks:
            outputs = self._get_intermediate_layers_chunked(x, n)
        else:
            outputs = self._get_intermediate_layers_not_chunked(x, n)
        if norm:
            outputs = [self.norm(out) for out in outputs]
        class_tokens = [out[:, 0] for out in outputs]
        outputs = [out[:, 1 + self.num_register_tokens :] for out in outputs]
        if reshape:
            B, _, w, h = x.shape
            outputs = [
                out.reshape(B, w // self.patch_size, h // self.patch_size, -1).permute(0, 3, 1, 2).contiguous()
                for out in outputs
            ]
        if return_class_token:
            return tuple(zip(outputs, class_tokens))
        return tuple(outputs)

    def forward(self, *args, is_training=False, **kwargs):
        ret = self.forward_features(*args, **kwargs)
        if is_training:
            return ret
        else:
            return self.head(ret["x_norm_clstoken"])


def init_weights_vit_timm(module: nn.Module, name: str = ""):
    """ViT weight initialization, original timm impl (for reproducibility)"""
    if isinstance(module, nn.Linear):
        trunc_normal_(module.weight, std=0.02)
        if module.bias is not None:
            nn.init.zeros_(module.bias)


def vit_small(patch_size=16, num_register_tokens=0, init_values=1.0, block_chunks=0,  **kwargs):
    model = DinoVisionTransformer(
        patch_size=patch_size,
        embed_dim=384,
        depth=12,
        num_heads=6,
        mlp_ratio=4,
        block_fn=partial(Block, attn_class=MemEffAttention),
        num_register_tokens=num_register_tokens,
        init_values=init_values,  # for layerscale: None or 0 => no layerscale
        block_chunks=block_chunks,
        **kwargs,
    )
    return model


def vit_base(patch_size=16, num_register_tokens=0, **kwargs):
    model = DinoVisionTransformer(
        patch_size=patch_size,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4,
        block_fn=partial(Block, attn_class=MemEffAttention),
        num_register_tokens=num_register_tokens,
        **kwargs,
    )
    return model


def vit_large(patch_size=16, num_register_tokens=0, **kwargs):
    model = DinoVisionTransformer(
        patch_size=patch_size,
        embed_dim=1024,
        depth=24,
        num_heads=16,
        mlp_ratio=4,
        block_fn=partial(Block, attn_class=MemEffAttention),
        num_register_tokens=num_register_tokens,
        **kwargs,
    )
    return model


def vit_giant2(patch_size=16, num_register_tokens=0, **kwargs):
    """
    Close to ViT-giant, with embed-dim 1536 and 24 heads => embed-dim per head 64
    """
    model = DinoVisionTransformer(
        patch_size=patch_size,
        embed_dim=1536,
        depth=40,
        num_heads=24,
        mlp_ratio=4,
        block_fn=partial(Block, attn_class=MemEffAttention),
        num_register_tokens=num_register_tokens,
        **kwargs,
    )
    return model