File size: 4,399 Bytes
ab687e7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import trunc_normal_

from ..swinv2_model import SwinTransformerV2


class SwinTransformerV2ForSimMIM(SwinTransformerV2):
    def __init__(self, **kwargs):
        super().__init__(**kwargs)

        assert self.num_classes == 0

        self.mask_token = nn.Parameter(torch.zeros(1, 1, self.embed_dim))
        trunc_normal_(self.mask_token, mean=0., std=.02)

    def forward(self, x, mask):
        x = self.patch_embed(x)

        assert mask is not None
        B, L, _ = x.shape

        mask_tokens = self.mask_token.expand(B, L, -1)
        w = mask.flatten(1).unsqueeze(-1).type_as(mask_tokens)
        x = x * (1. - w) + mask_tokens * w

        if self.ape:
            x = x + self.absolute_pos_embed
        x = self.pos_drop(x)

        for layer in self.layers:
            x = layer(x)
        x = self.norm(x)

        x = x.transpose(1, 2)
        B, C, L = x.shape
        H = W = int(L ** 0.5)
        x = x.reshape(B, C, H, W)
        return x

    @torch.jit.ignore
    def no_weight_decay(self):
        return super().no_weight_decay() | {'mask_token'}


class MiMModel(nn.Module):
    """
    Masked-Image-Modeling model

    Given an encoder, makes a model that incorporates
    the encoder and attaches a simple linear layer that
    produces the raw-pixel predictions of the masked
    inputs.
    """
    def __init__(self, encoder, encoder_stride, in_chans, patch_size):
        super().__init__()
        self.encoder = encoder
        self.encoder_stride = encoder_stride
        self.in_chans = in_chans
        self.patch_size = patch_size
        self.decoder = nn.Sequential(
            nn.Conv2d(
                in_channels=self.encoder.num_features,
                out_channels=self.encoder_stride ** 2 * self.in_chans,
                kernel_size=1),
            nn.PixelShuffle(self.encoder_stride),
        )

        # self.in_chans = self.encoder.in_chans
        # self.patch_size = self.encoder.patch_size

    def forward(self, x, mask):
        z = self.encoder(x, mask)
        x_rec = self.decoder(z)

        mask = mask.repeat_interleave(self.patch_size, 1).repeat_interleave(
            self.patch_size, 2).unsqueeze(1).contiguous()
        loss_recon = F.l1_loss(x, x_rec, reduction='none')
        loss = (loss_recon * mask).sum() / (mask.sum() + 1e-5) / self.in_chans
        return loss

    @torch.jit.ignore
    def no_weight_decay(self):
        if hasattr(self.encoder, 'no_weight_decay'):
            return {'encoder.' + i for i in self.encoder.no_weight_decay()}
        return {}

    @torch.jit.ignore
    def no_weight_decay_keywords(self):
        if hasattr(self.encoder, 'no_weight_decay_keywords'):
            return {'encoder.' + i for i in
                    self.encoder.no_weight_decay_keywords()}
        return {}


def build_mim_model(config):
    """Builds the masked-image-modeling model.

    Args:
        config: config object

    Raises:
        NotImplementedError: if the model is
        not swinv2, then this will be thrown.

    Returns:
        MiMModel: masked-image-modeling model
    """
    model_type = config.MODEL.TYPE
    if model_type == 'swinv2':
        encoder = SwinTransformerV2ForSimMIM(
            img_size=config.DATA.IMG_SIZE,
            patch_size=config.MODEL.SWINV2.PATCH_SIZE,
            in_chans=config.MODEL.SWINV2.IN_CHANS,
            num_classes=0,
            embed_dim=config.MODEL.SWINV2.EMBED_DIM,
            depths=config.MODEL.SWINV2.DEPTHS,
            num_heads=config.MODEL.SWINV2.NUM_HEADS,
            window_size=config.MODEL.SWINV2.WINDOW_SIZE,
            mlp_ratio=config.MODEL.SWINV2.MLP_RATIO,
            qkv_bias=config.MODEL.SWINV2.QKV_BIAS,
            drop_rate=config.MODEL.DROP_RATE,
            drop_path_rate=config.MODEL.DROP_PATH_RATE,
            ape=config.MODEL.SWINV2.APE,
            patch_norm=config.MODEL.SWINV2.PATCH_NORM,
            use_checkpoint=config.TRAIN.USE_CHECKPOINT)
        encoder_stride = 32
        in_chans = config.MODEL.SWINV2.IN_CHANS
        patch_size = config.MODEL.SWINV2.PATCH_SIZE
    else:
        raise NotImplementedError(f"Unknown pre-train model: {model_type}")

    model = MiMModel(encoder=encoder, encoder_stride=encoder_stride,
                     in_chans=in_chans, patch_size=patch_size)

    return model