【图像分类】2021-CvT

论文题目：CvT: Introducing Convolutions to Vision Transformers

论文链接：https://arxiv.org/abs/2103.15808

论文代码：https://github.com/microsoft/CvT

1. 简介

1.1 简介

方法简洁高效，性能在现在大神云集的Transformer算法里非常有竞争力。

2. 网络

2.1 整体架构

作者提出了三种大小的网络，CvT13 CvT21和CvTW24

2.2 Convolutional Token Embedding

通过卷积7*7获得conv embedding。conv2d并不是深度卷积，可以理解为一个patch embedding操作，在大部分cv transformer中均会进行。这里是每个阶段会使用一次Patch embedding,来增加通道和降低分辨率。

import torch
import torch.nn as nn
import torch.nn.functional as F
from einops import rearrange
from einops.layers.torch import Rearrange
from timm.models.layers import DropPath, trunc_normal_, to_2tuple

class ConvEmbed(nn.Module):
    """ Image to Conv Embedding
    """

    def __init__(self,
                 patch_size=7,
                 in_chans=3,
                 embed_dim=64,
                 stride=4,
                 padding=2,
                 norm_layer=None):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        self.patch_size = patch_size

        self.proj = nn.Conv2d(
            in_chans, embed_dim,
            kernel_size=patch_size,
            stride=stride,
            padding=padding
        )
        self.norm = norm_layer(embed_dim) if norm_layer else None

    def forward(self, x):
        x = self.proj(x)

        B, C, H, W = x.shape
        x = rearrange(x, 'b c h w -> b (h w) c')
        if self.norm:
            x = self.norm(x)
        x = rearrange(x, 'b (h w) c -> b c h w', h=H, w=W)

        return x
if __name__ == '__main__':
    x=torch.randn(1,3,224,224)
    model=ConvEmbed()
    y=model(x)
    #  patch_size=7,embed_dim=64,stride=4
    #  224*224 的图片做 4倍下采样。变成 56*56
    # 
    print(y.shape)

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

可以看到，输入的x经过卷积后，生成特征图f（b,c w,h）并将其压缩成(b hw c)的形状放入LayerNorm中计算，最后在变回去f(b,c,w,h) 用于下一层的计算。

其实这里的f(b hw c)就是Token，其中HW就是Token的个数，c是Token的维度。

相比起ViT，ConvEmbed没有将Patch_size和stride等同，也就是说，卷积会有很多重叠的部分，生成的Token个数要远远多于ViT的Patch_size*Patch_size

官方实现给出的第一层Patch_size=(7,7) stride=4

原本ViT的patch embedding

import torch
import torch.nn as nn
import torch.nn.functional as F
from einops import rearrange
from einops.layers.torch import Rearrange
from timm.models.layers import DropPath, trunc_normal_, to_2tuple


class PatchEmbed(nn.Module):
    def __init__(self, patch_size=7, channels=3, embed_dim=64, norm_layer=None):
        super(PatchEmbed, self).__init__()
        patch_height = patch_size
        patch_width = patch_height
        patch_dim = channels * patch_height * patch_width
        self.to_patch_embedding = nn.Sequential(
            Rearrange('b c (h p1) (w p2) -> b (h w) (p1 p2 c)', p1=patch_height, p2=patch_width),
            nn.Linear(patch_dim, embed_dim),
        )

    def forward(self, x):
        return self.to_patch_embedding(x)

if __name__ == '__main__':
    x=torch.randn(1,3,224,224)
    model=PatchEmbed()
    y=model(x)
    #  patch_size=7,embed_dim=64,stride=4
    #  224*224 的图片做 4倍下采样。变成 56*56
    #
    print(y.shape)


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

2.3 Convolutional Projection for Attention(使用卷积生成kqv)

通过深度卷积进行conv proj，即将特征转化成query ，value,key向量。这种转化方式可以见下

• (a)是ViT中生成QKV的方法
• (b)是卷积生成QKV的方法
• ©是使用Squeezed Convolution生成QKV的方法

def forward_conv(self, x, h, w):
    if self.with_cls_token:
        cls_token, x = torch.split(x, [1, h * w], 1)

    x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w)

    if self.conv_proj_q is not None:
        q = self.conv_proj_q(x)
    else:
        q = rearrange(x, 'b c h w -> b (h w) c')

    if self.conv_proj_k is not None:
        k = self.conv_proj_k(x)
    else:
        k = rearrange(x, 'b c h w -> b (h w) c')

    if self.conv_proj_v is not None:
        v = self.conv_proj_v(x)
    else:
        v = rearrange(x, 'b c h w -> b (h w) c')

    if self.with_cls_token:
        q = torch.cat((cls_token, q), dim=1)
        k = torch.cat((cls_token, k), dim=1)
        v = torch.cat((cls_token, v), dim=1)

    return q, k, v
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

传入的x(b hw c)，即上一层产生的Token，首先将其解压缩变形成(b c h w)，然后分别使用三种卷积核conv_proj_q、conv_proj_k、conv_proj_v生成QKV然后加上cls_token。

其中卷积操作是Depth-wise Convolution+BatchNorm实现的

如图b所示，直接使用卷积操作，需要的参数量是$s^2{C}^2$,复杂度是$O(s^2{C}^{2}T)$

• s是kernel size
• C是 Token Channel Dimension
• T是 Token number

为了减少参数量，这里使用Depth-wise separeble convolution。这样参数量就减少到了$s^{2}C$,复杂度就减少到了$O(s^{2}CT)$

与此同时，在生成KV的时候，将步长调整到2或者更大，可以减少KV的尺寸，生成Local Sequeeze Key和Local Sequeeze Value

图像中相邻像素会在外观和语义上有信息冗余，这样做会引起微小的性能损失，但是Q补偿了这部分损失。

2.4 移除位置信息

在这个框架中，Convolution Projection和Convolutional Token Embedding应用在每一个Transformer Block中。卷积本身就带有位置信息，给予了模型感知位置信息的能力。因此可以将位置编码安全的移除并不影响任何性能损失。

在实验部分，作者做了消融实验，结果如下：

3. 代码

3.1 原版

from functools import partial
from itertools import repeat

import logging
import os
from collections import OrderedDict

import numpy as np
import scipy
import torch
import torch.nn as nn
import torch.nn.functional as F
from einops import rearrange
from einops.layers.torch import Rearrange

from timm.models.layers import DropPath, trunc_normal_, to_2tuple


# # From PyTorch internals
# def _ntuple(n):
#     def parse(x):
#         if isinstance(x, container_abcs.Iterable):
#             return x
#         return tuple(repeat(x, n))
#
#     return parse

#
# to_1tuple = _ntuple(1)
# to_2tuple = _ntuple(2)
# to_3tuple = _ntuple(3)
# to_4tuple = _ntuple(4)
# to_ntuple = _ntuple


class LayerNorm(nn.LayerNorm):
    """Subclass torch's LayerNorm to handle fp16."""

    def forward(self, x: torch.Tensor):
        orig_type = x.dtype
        ret = super().forward(x.type(torch.float32))
        return ret.type(orig_type)


class QuickGELU(nn.Module):
    def forward(self, x: torch.Tensor):
        return x * torch.sigmoid(1.702 * x)


class Mlp(nn.Module):
    def __init__(self,
                 in_features,
                 hidden_features=None,
                 out_features=None,
                 act_layer=nn.GELU,
                 drop=0.):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.fc1 = nn.Linear(in_features, hidden_features)
        self.act = act_layer()
        self.fc2 = nn.Linear(hidden_features, out_features)
        self.drop = nn.Dropout(drop)

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


class Attention(nn.Module):
    def __init__(self,
                 dim_in,
                 dim_out,
                 num_heads,
                 qkv_bias=False,
                 attn_drop=0.,
                 proj_drop=0.,
                 method='dw_bn',
                 kernel_size=3,
                 stride_kv=1,
                 stride_q=1,
                 padding_kv=1,
                 padding_q=1,
                 with_cls_token=True,
                 **kwargs
                 ):
        super().__init__()
        self.stride_kv = stride_kv
        self.stride_q = stride_q
        self.dim = dim_out
        self.num_heads = num_heads
        # head_dim = self.qkv_dim // num_heads
        self.scale = dim_out ** -0.5
        self.with_cls_token = with_cls_token

        self.conv_proj_q = self._build_projection(
            dim_in, dim_out, kernel_size, padding_q,
            stride_q, 'linear' if method == 'avg' else method
        )
        self.conv_proj_k = self._build_projection(
            dim_in, dim_out, kernel_size, padding_kv,
            stride_kv, method
        )
        self.conv_proj_v = self._build_projection(
            dim_in, dim_out, kernel_size, padding_kv,
            stride_kv, method
        )

        self.proj_q = nn.Linear(dim_in, dim_out, bias=qkv_bias)
        self.proj_k = nn.Linear(dim_in, dim_out, bias=qkv_bias)
        self.proj_v = nn.Linear(dim_in, dim_out, bias=qkv_bias)

        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim_out, dim_out)
        self.proj_drop = nn.Dropout(proj_drop)

    def _build_projection(self,
                          dim_in,
                          dim_out,
                          kernel_size,
                          padding,
                          stride,
                          method):
        if method == 'dw_bn':
            proj = nn.Sequential(OrderedDict([
                ('conv', nn.Conv2d(
                    dim_in,
                    dim_in,
                    kernel_size=kernel_size,
                    padding=padding,
                    stride=stride,
                    bias=False,
                    groups=dim_in
                )),
                ('bn', nn.BatchNorm2d(dim_in)),
                ('rearrage', Rearrange('b c h w -> b (h w) c')),
            ]))

        elif method == 'avg':
            proj = nn.Sequential(OrderedDict([
                ('avg', nn.AvgPool2d(
                    kernel_size=kernel_size,
                    padding=padding,
                    stride=stride,
                    ceil_mode=True
                )),
                ('rearrage', Rearrange('b c h w -> b (h w) c')),
            ]))
        elif method == 'linear':
            proj = None
        else:
            raise ValueError('Unknown method ({})'.format(method))

        return proj

    def forward_conv(self, x, h, w):
        if self.with_cls_token:
            cls_token, x = torch.split(x, [1, h * w], 1)

        x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w)

        if self.conv_proj_q is not None:
            q = self.conv_proj_q(x)
        else:
            q = rearrange(x, 'b c h w -> b (h w) c')

        if self.conv_proj_k is not None:
            k = self.conv_proj_k(x)
        else:
            k = rearrange(x, 'b c h w -> b (h w) c')

        if self.conv_proj_v is not None:
            v = self.conv_proj_v(x)
        else:
            v = rearrange(x, 'b c h w -> b (h w) c')

        if self.with_cls_token:
            q = torch.cat((cls_token, q), dim=1)
            k = torch.cat((cls_token, k), dim=1)
            v = torch.cat((cls_token, v), dim=1)

        return q, k, v

    def forward(self, x, h, w):
        if (
                self.conv_proj_q is not None
                or self.conv_proj_k is not None
                or self.conv_proj_v is not None
        ):
            q, k, v = self.forward_conv(x, h, w)

        q = rearrange(self.proj_q(q), 'b t (h d) -> b h t d', h=self.num_heads)
        k = rearrange(self.proj_k(k), 'b t (h d) -> b h t d', h=self.num_heads)
        v = rearrange(self.proj_v(v), 'b t (h d) -> b h t d', h=self.num_heads)

        attn_score = torch.einsum('bhlk,bhtk->bhlt', [q, k]) * self.scale
        attn = F.softmax(attn_score, dim=-1)
        attn = self.attn_drop(attn)

        x = torch.einsum('bhlt,bhtv->bhlv', [attn, v])
        x = rearrange(x, 'b h t d -> b t (h d)')

        x = self.proj(x)
        x = self.proj_drop(x)

        return x

    @staticmethod
    def compute_macs(module, input, output):
        # T: num_token
        # S: num_token
        input = input[0]
        flops = 0

        _, T, C = input.shape
        H = W = int(np.sqrt(T - 1)) if module.with_cls_token else int(np.sqrt(T))

        H_Q = H / module.stride_q
        W_Q = H / module.stride_q
        T_Q = H_Q * W_Q + 1 if module.with_cls_token else H_Q * W_Q

        H_KV = H / module.stride_kv
        W_KV = W / module.stride_kv
        T_KV = H_KV * W_KV + 1 if module.with_cls_token else H_KV * W_KV

        # C = module.dim
        # S = T
        # Scaled-dot-product macs
        # [B x T x C] x [B x C x T] --> [B x T x S]
        # multiplication-addition is counted as 1 because operations can be fused
        flops += T_Q * T_KV * module.dim
        # [B x T x S] x [B x S x C] --> [B x T x C]
        flops += T_Q * module.dim * T_KV

        if (
                hasattr(module, 'conv_proj_q')
                and hasattr(module.conv_proj_q, 'conv')
        ):
            params = sum(
                [
                    p.numel()
                    for p in module.conv_proj_q.conv.parameters()
                ]
            )
            flops += params * H_Q * W_Q

        if (
                hasattr(module, 'conv_proj_k')
                and hasattr(module.conv_proj_k, 'conv')
        ):
            params = sum(
                [
                    p.numel()
                    for p in module.conv_proj_k.conv.parameters()
                ]
            )
            flops += params * H_KV * W_KV

        if (
                hasattr(module, 'conv_proj_v')
                and hasattr(module.conv_proj_v, 'conv')
        ):
            params = sum(
                [
                    p.numel()
                    for p in module.conv_proj_v.conv.parameters()
                ]
            )
            flops += params * H_KV * W_KV

        params = sum([p.numel() for p in module.proj_q.parameters()])
        flops += params * T_Q
        params = sum([p.numel() for p in module.proj_k.parameters()])
        flops += params * T_KV
        params = sum([p.numel() for p in module.proj_v.parameters()])
        flops += params * T_KV
        params = sum([p.numel() for p in module.proj.parameters()])
        flops += params * T

        module.__flops__ += flops


class Block(nn.Module):

    def __init__(self,
                 dim_in,
                 dim_out,
                 num_heads,
                 mlp_ratio=4.,
                 qkv_bias=False,
                 drop=0.,
                 attn_drop=0.,
                 drop_path=0.,
                 act_layer=nn.GELU,
                 norm_layer=nn.LayerNorm,
                 **kwargs):
        super().__init__()

        self.with_cls_token = kwargs['with_cls_token']

        self.norm1 = norm_layer(dim_in)
        self.attn = Attention(
            dim_in, dim_out, num_heads, qkv_bias, attn_drop, drop,
            **kwargs
        )

        self.drop_path = DropPath(drop_path) \
            if drop_path > 0. else nn.Identity()
        self.norm2 = norm_layer(dim_out)

        dim_mlp_hidden = int(dim_out * mlp_ratio)
        self.mlp = Mlp(
            in_features=dim_out,
            hidden_features=dim_mlp_hidden,
            act_layer=act_layer,
            drop=drop
        )

    def forward(self, x, h, w):
        res = x

        x = self.norm1(x)
        attn = self.attn(x, h, w)
        x = res + self.drop_path(attn)
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x


class ConvEmbed(nn.Module):
    """ Image to Conv Embedding
    """

    def __init__(self,
                 patch_size=7,
                 in_chans=3,
                 embed_dim=64,
                 stride=4,
                 padding=2,
                 norm_layer=None):
        super().__init__()
        patch_size = to_2tuple(patch_size)
        self.patch_size = patch_size

        self.proj = nn.Conv2d(
            in_chans, embed_dim,
            kernel_size=patch_size,
            stride=stride,
            padding=padding
        )
        self.norm = norm_layer(embed_dim) if norm_layer else None

    def forward(self, x):
        x = self.proj(x)

        B, C, H, W = x.shape
        x = rearrange(x, 'b c h w -> b (h w) c')
        if self.norm:
            x = self.norm(x)
        x = rearrange(x, 'b (h w) c -> b c h w', h=H, w=W)

        return x


class VisionTransformer(nn.Module):
    """ Vision Transformer with support for patch or hybrid CNN input stage
    """

    def __init__(self,
                 patch_size=16,
                 patch_stride=16,
                 patch_padding=0,
                 in_chans=3,
                 embed_dim=768,
                 depth=12,
                 num_heads=12,
                 mlp_ratio=4.,
                 qkv_bias=False,
                 drop_rate=0.,
                 attn_drop_rate=0.,
                 drop_path_rate=0.,
                 act_layer=nn.GELU,
                 norm_layer=nn.LayerNorm,
                 init='trunc_norm',
                 **kwargs):
        super().__init__()
        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models

        self.rearrage = None

        self.patch_embed = ConvEmbed(
            # img_size=img_size,
            patch_size=patch_size,
            in_chans=in_chans,
            stride=patch_stride,
            padding=patch_padding,
            embed_dim=embed_dim,
            norm_layer=norm_layer
        )

        with_cls_token = kwargs['with_cls_token']
        if with_cls_token:
            self.cls_token = nn.Parameter(
                torch.zeros(1, 1, embed_dim)
            )
        else:
            self.cls_token = None

        self.pos_drop = nn.Dropout(p=drop_rate)
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule

        blocks = []
        for j in range(depth):
            blocks.append(
                Block(
                    dim_in=embed_dim,
                    dim_out=embed_dim,
                    num_heads=num_heads,
                    mlp_ratio=mlp_ratio,
                    qkv_bias=qkv_bias,
                    drop=drop_rate,
                    attn_drop=attn_drop_rate,
                    drop_path=dpr[j],
                    act_layer=act_layer,
                    norm_layer=norm_layer,
                    **kwargs
                )
            )
        self.blocks = nn.ModuleList(blocks)

        if self.cls_token is not None:
            trunc_normal_(self.cls_token, std=.02)

        if init == 'xavier':
            self.apply(self._init_weights_xavier)
        else:
            self.apply(self._init_weights_trunc_normal)

    def _init_weights_trunc_normal(self, m):
        if isinstance(m, nn.Linear):
            logging.info('=> init weight of Linear from trunc norm')
            trunc_normal_(m.weight, std=0.02)
            if m.bias is not None:
                logging.info('=> init bias of Linear to zeros')
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, (nn.LayerNorm, nn.BatchNorm2d)):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    def _init_weights_xavier(self, m):
        if isinstance(m, nn.Linear):
            logging.info('=> init weight of Linear from xavier uniform')
            nn.init.xavier_uniform_(m.weight)
            if m.bias is not None:
                logging.info('=> init bias of Linear to zeros')
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, (nn.LayerNorm, nn.BatchNorm2d)):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    def forward(self, x):
        x = self.patch_embed(x)
        B, C, H, W = x.size()

        x = rearrange(x, 'b c h w -> b (h w) c')

        cls_tokens = None
        if self.cls_token is not None:
            # stole cls_tokens impl from Phil Wang, thanks
            cls_tokens = self.cls_token.expand(B, -1, -1)
            x = torch.cat((cls_tokens, x), dim=1)

        x = self.pos_drop(x)

        for i, blk in enumerate(self.blocks):
            x = blk(x, H, W)

        if self.cls_token is not None:
            cls_tokens, x = torch.split(x, [1, H * W], 1)
        x = rearrange(x, 'b (h w) c -> b c h w', h=H, w=W)

        return x, cls_tokens


class ConvolutionalVisionTransformer(nn.Module):
    def __init__(self,
                 in_chans=3,
                 num_classes=1000,
                 act_layer=nn.GELU,
                 norm_layer=nn.LayerNorm,
                 init='trunc_norm',
                 spec=None):
        super().__init__()
        self.num_classes = num_classes

        self.num_stages = spec['NUM_STAGES']
        for i in range(self.num_stages):
            kwargs = {
                'patch_size': spec['PATCH_SIZE'][i],
                'patch_stride': spec['PATCH_STRIDE'][i],
                'patch_padding': spec['PATCH_PADDING'][i],
                'embed_dim': spec['DIM_EMBED'][i],
                'depth': spec['DEPTH'][i],
                'num_heads': spec['NUM_HEADS'][i],
                'mlp_ratio': spec['MLP_RATIO'][i],
                'qkv_bias': spec['QKV_BIAS'][i],
                'drop_rate': spec['DROP_RATE'][i],
                'attn_drop_rate': spec['ATTN_DROP_RATE'][i],
                'drop_path_rate': spec['DROP_PATH_RATE'][i],
                'with_cls_token': spec['CLS_TOKEN'][i],
                'method': spec['QKV_PROJ_METHOD'][i],
                'kernel_size': spec['KERNEL_QKV'][i],
                'padding_q': spec['PADDING_Q'][i],
                'padding_kv': spec['PADDING_KV'][i],
                'stride_kv': spec['STRIDE_KV'][i],
                'stride_q': spec['STRIDE_Q'][i],
            }

            stage = VisionTransformer(
                in_chans=in_chans,
                init=init,
                act_layer=act_layer,
                norm_layer=norm_layer,
                **kwargs
            )
            setattr(self, f'stage{i}', stage)

            in_chans = spec['DIM_EMBED'][i]

        dim_embed = spec['DIM_EMBED'][-1]
        self.norm = norm_layer(dim_embed)
        self.cls_token = spec['CLS_TOKEN'][-1]

        # Classifier head
        self.head = nn.Linear(dim_embed, num_classes) if num_classes > 0 else nn.Identity()
        trunc_normal_(self.head.weight, std=0.02)

    def init_weights(self, pretrained='', pretrained_layers=[], verbose=True):
        if os.path.isfile(pretrained):
            pretrained_dict = torch.load(pretrained, map_location='cpu')
            logging.info(f'=> loading pretrained model {pretrained}')
            model_dict = self.state_dict()
            pretrained_dict = {
                k: v for k, v in pretrained_dict.items()
                if k in model_dict.keys()
            }
            need_init_state_dict = {}
            for k, v in pretrained_dict.items():
                need_init = (
                        k.split('.')[0] in pretrained_layers
                        or pretrained_layers[0] is '*'
                )
                if need_init:
                    if verbose:
                        logging.info(f'=> init {k} from {pretrained}')
                    if 'pos_embed' in k and v.size() != model_dict[k].size():
                        size_pretrained = v.size()
                        size_new = model_dict[k].size()
                        logging.info(
                            '=> load_pretrained: resized variant: {} to {}'
                                .format(size_pretrained, size_new)
                        )

                        ntok_new = size_new[1]
                        ntok_new -= 1

                        posemb_tok, posemb_grid = v[:, :1], v[0, 1:]

                        gs_old = int(np.sqrt(len(posemb_grid)))
                        gs_new = int(np.sqrt(ntok_new))

                        logging.info(
                            '=> load_pretrained: grid-size from {} to {}'
                                .format(gs_old, gs_new)
                        )

                        posemb_grid = posemb_grid.reshape(gs_old, gs_old, -1)
                        zoom = (gs_new / gs_old, gs_new / gs_old, 1)
                        posemb_grid = scipy.ndimage.zoom(
                            posemb_grid, zoom, order=1
                        )
                        posemb_grid = posemb_grid.reshape(1, gs_new ** 2, -1)
                        v = torch.tensor(
                            np.concatenate([posemb_tok, posemb_grid], axis=1)
                        )

                    need_init_state_dict[k] = v
            self.load_state_dict(need_init_state_dict, strict=False)

    @torch.jit.ignore
    def no_weight_decay(self):
        layers = set()
        for i in range(self.num_stages):
            layers.add(f'stage{i}.pos_embed')
            layers.add(f'stage{i}.cls_token')

        return layers

    def forward_features(self, x):
        for i in range(self.num_stages):
            x, cls_tokens = getattr(self, f'stage{i}')(x)

        if self.cls_token:
            x = self.norm(cls_tokens)
            x = torch.squeeze(x)
        else:
            x = rearrange(x, 'b c h w -> b (h w) c')
            x = self.norm(x)
            x = torch.mean(x, dim=1)
        return x

    def forward(self, x):
        x = self.forward_features(x)
        # print(x.shape)
        x = self.head(x)
        return x


def CvT_13(**kwargs):
    spec = {
        "INIT": 'trunc_norm',
        "NUM_STAGES": 3,
        "PATCH_SIZE": [7, 3, 3],
        "PATCH_STRIDE": [4, 2, 2],
        "PATCH_PADDING": [2, 1, 1],
        "DIM_EMBED": [64, 192, 384],
        "NUM_HEADS": [1, 3, 6],
        "DEPTH": [1, 2, 10],
        "MLP_RATIO": [4.0, 4.0, 4.0],
        "ATTN_DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_PATH_RATE": [0.0, 0.0, 0.1],
        "QKV_BIAS": [True, True, True],
        "CLS_TOKEN": [False, False, True],
        "POS_EMBED": [False, False, False],
        "QKV_PROJ_METHOD": ['dw_bn', 'dw_bn', 'dw_bn'],
        "KERNEL_QKV": [3, 3, 3],
        "PADDING_KV": [1, 1, 1],
        "STRIDE_KV": [2, 2, 2],
        "PADDING_Q": [1, 1, 1],
        "STRIDE_Q": [1, 1, 1],
    }
    model = ConvolutionalVisionTransformer(
        **kwargs,
        spec=spec,
    )
    return model


def CvT_21(**kwargs):
    spec = {
        "INIT": 'trunc_norm',
        "NUM_STAGES": 3,
        "PATCH_SIZE": [7, 3, 3],
        "PATCH_STRIDE": [4, 2, 2],
        "PATCH_PADDING": [2, 1, 1],
        "DIM_EMBED": [64, 192, 384],
        "NUM_HEADS": [1, 3, 6],
        "DEPTH": [1, 4, 16],
        "MLP_RATIO": [4.0, 4.0, 4.0],
        "ATTN_DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_PATH_RATE": [0.0, 0.0, 0.1],
        "QKV_BIAS": [True, True, True],
        "CLS_TOKEN": [False, False, True],
        "POS_EMBED": [False, False, False],
        "QKV_PROJ_METHOD": ['dw_bn', 'dw_bn', 'dw_bn'],
        "KERNEL_QKV": [3, 3, 3],
        "PADDING_KV": [1, 1, 1],
        "STRIDE_KV": [2, 2, 2],
        "PADDING_Q": [1, 1, 1],
        "STRIDE_Q": [1, 1, 1],
    }
    model = ConvolutionalVisionTransformer(
        **kwargs,
        spec=spec,
    )
    return model


def CvT_W24(**kwargs):
    spec = {
        "INIT": 'trunc_norm',
        "NUM_STAGES": 3,
        "PATCH_SIZE": [7, 3, 3],
        "PATCH_STRIDE": [4, 2, 2],
        "PATCH_PADDING": [2, 1, 1],
        "DIM_EMBED": [192, 768, 1024],
        "NUM_HEADS": [3, 12, 16],
        "DEPTH": [2, 2, 20],
        "MLP_RATIO": [4.0, 4.0, 4.0],
        "ATTN_DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_RATE": [0.0, 0.0, 0.0],
        "DROP_PATH_RATE": [0.0, 0.0, 0.3],
        "QKV_BIAS": [True, True, True],
        "CLS_TOKEN": [False, False, True],
        "POS_EMBED": [False, False, False],
        "QKV_PROJ_METHOD": ['dw_bn', 'dw_bn', 'dw_bn'],
        "KERNEL_QKV": [3, 3, 3],
        "PADDING_KV": [1, 1, 1],
        "STRIDE_KV": [2, 2, 2],
        "PADDING_Q": [1, 1, 1],
        "STRIDE_Q": [1, 1, 1],
    }
    model = ConvolutionalVisionTransformer(
        **kwargs,
        spec=spec,
    )
    return model


if __name__ == '__main__':
    x = torch.randn(2, 3, 224, 224)
    model = CvT_13(in_chans=3,
                   num_classes=1000,
                   act_layer=QuickGELU,
                   norm_layer=partial(LayerNorm, eps=1e-5))
    y = model(x)
    print(y.shape)

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
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724

3.2 简略版

import torch
from torch import nn, einsum
import torch.nn.functional as F

from einops import rearrange, repeat
from einops.layers.torch import Rearrange

# helper methods

def group_dict_by_key(cond, d):
    return_val = [dict(), dict()]
    for key in d.keys():
        match = bool(cond(key))
        ind = int(not match)
        return_val[ind][key] = d[key]
    return (*return_val,)

def group_by_key_prefix_and_remove_prefix(prefix, d):
    kwargs_with_prefix, kwargs = group_dict_by_key(lambda x: x.startswith(prefix), d)
    kwargs_without_prefix = dict(map(lambda x: (x[0][len(prefix):], x[1]), tuple(kwargs_with_prefix.items())))
    return kwargs_without_prefix, kwargs

# classes

class LayerNorm(nn.Module): # layernorm, but done in the channel dimension #1
    def __init__(self, dim, eps = 1e-5):
        super().__init__()
        self.eps = eps
        self.g = nn.Parameter(torch.ones(1, dim, 1, 1))
        self.b = nn.Parameter(torch.zeros(1, dim, 1, 1))

    def forward(self, x):
        var = torch.var(x, dim = 1, unbiased = False, keepdim = True)
        mean = torch.mean(x, dim = 1, keepdim = True)
        return (x - mean) / (var + self.eps).sqrt() * self.g + self.b

class PreNorm(nn.Module):
    def __init__(self, dim, fn):
        super().__init__()
        self.norm = LayerNorm(dim)
        self.fn = fn
    def forward(self, x, **kwargs):
        x = self.norm(x)
        return self.fn(x, **kwargs)

class FeedForward(nn.Module):
    def __init__(self, dim, mult = 4, dropout = 0.):
        super().__init__()
        self.net = nn.Sequential(
            nn.Conv2d(dim, dim * mult, 1),
            nn.GELU(),
            nn.Dropout(dropout),
            nn.Conv2d(dim * mult, dim, 1),
            nn.Dropout(dropout)
        )
    def forward(self, x):
        return self.net(x)

class DepthWiseConv2d(nn.Module):
    def __init__(self, dim_in, dim_out, kernel_size, padding, stride, bias = True):
        super().__init__()
        self.net = nn.Sequential(
            nn.Conv2d(dim_in, dim_in, kernel_size = kernel_size, padding = padding, groups = dim_in, stride = stride, bias = bias),
            nn.BatchNorm2d(dim_in),
            nn.Conv2d(dim_in, dim_out, kernel_size = 1, bias = bias)
        )
    def forward(self, x):
        return self.net(x)

class Attention(nn.Module):
    def __init__(self, dim, proj_kernel, kv_proj_stride, heads = 8, dim_head = 64, dropout = 0.):
        super().__init__()
        inner_dim = dim_head *  heads
        padding = proj_kernel // 2
        self.heads = heads
        self.scale = dim_head ** -0.5

        self.attend = nn.Softmax(dim = -1)
        self.dropout = nn.Dropout(dropout)

        self.to_q = DepthWiseConv2d(dim, inner_dim, proj_kernel, padding = padding, stride = 1, bias = False)
        self.to_kv = DepthWiseConv2d(dim, inner_dim * 2, proj_kernel, padding = padding, stride = kv_proj_stride, bias = False)

        self.to_out = nn.Sequential(
            nn.Conv2d(inner_dim, dim, 1),
            nn.Dropout(dropout)
        )

    def forward(self, x):
        shape = x.shape
        b, n, _, y, h = *shape, self.heads
        q, k, v = (self.to_q(x), *self.to_kv(x).chunk(2, dim = 1))
        q, k, v = map(lambda t: rearrange(t, 'b (h d) x y -> (b h) (x y) d', h = h), (q, k, v))

        dots = einsum('b i d, b j d -> b i j', q, k) * self.scale

        attn = self.attend(dots)
        attn = self.dropout(attn)

        out = einsum('b i j, b j d -> b i d', attn, v)
        out = rearrange(out, '(b h) (x y) d -> b (h d) x y', h = h, y = y)
        return self.to_out(out)

class Transformer(nn.Module):
    def __init__(self, dim, proj_kernel, kv_proj_stride, depth, heads, dim_head = 64, mlp_mult = 4, dropout = 0.):
        super().__init__()
        self.layers = nn.ModuleList([])
        for _ in range(depth):
            self.layers.append(nn.ModuleList([
                PreNorm(dim, Attention(dim, proj_kernel = proj_kernel, kv_proj_stride = kv_proj_stride, heads = heads, dim_head = dim_head, dropout = dropout)),
                PreNorm(dim, FeedForward(dim, mlp_mult, dropout = dropout))
            ]))
    def forward(self, x):
        for attn, ff in self.layers:
            x = attn(x) + x
            x = ff(x) + x
        return x

class CvT(nn.Module):
    def __init__(
        self,
        *,
        num_classes,
        s1_emb_dim = 64,
        s1_emb_kernel = 7,
        s1_emb_stride = 4,
        s1_proj_kernel = 3,
        s1_kv_proj_stride = 2,
        s1_heads = 1,
        s1_depth = 1,
        s1_mlp_mult = 4,
        s2_emb_dim = 192,
        s2_emb_kernel = 3,
        s2_emb_stride = 2,
        s2_proj_kernel = 3,
        s2_kv_proj_stride = 2,
        s2_heads = 3,
        s2_depth = 2,
        s2_mlp_mult = 4,
        s3_emb_dim = 384,
        s3_emb_kernel = 3,
        s3_emb_stride = 2,
        s3_proj_kernel = 3,
        s3_kv_proj_stride = 2,
        s3_heads = 6,
        s3_depth = 10,
        s3_mlp_mult = 4,
        dropout = 0.
    ):
        super().__init__()
        kwargs = dict(locals())

        dim = 3
        layers = []

        for prefix in ('s1', 's2', 's3'):
            config, kwargs = group_by_key_prefix_and_remove_prefix(f'{prefix}_', kwargs)

            layers.append(nn.Sequential(
                nn.Conv2d(dim, config['emb_dim'], kernel_size = config['emb_kernel'], padding = (config['emb_kernel'] // 2), stride = config['emb_stride']),
                LayerNorm(config['emb_dim']),
                Transformer(dim = config['emb_dim'], proj_kernel = config['proj_kernel'], kv_proj_stride = config['kv_proj_stride'], depth = config['depth'], heads = config['heads'], mlp_mult = config['mlp_mult'], dropout = dropout)
            ))

            dim = config['emb_dim']

        self.layers = nn.Sequential(*layers)

        self.to_logits = nn.Sequential(
            nn.AdaptiveAvgPool2d(1),
            Rearrange('... () () -> ...'),
            nn.Linear(dim, num_classes)
        )

    def forward(self, x):
        latents = self.layers(x)
        return self.to_logits(latents)
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

参考链接

【论文笔记】CvT: Introducing Convolutions to Vision Transformers_来自γ星的赛亚人的博客-CSDN博客