說之前先提一個視頻這個視頻還是很好的將transformer機(jī)制的變遷及未來的趨勢很詳細(xì)的說明了一下我覺得蠻有感觸的，建議可以看看這里首先提一下代碼及其對應(yīng)的論文視頻地址蚌卤。
paper：Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
code： microsoft/Swin-Transformer
可以理解SwinTransformer是新一代的特征提取神器奶镶，很多榜單都有它的影子羹奉，這里我們可以理解為是一種新的`backbone鲫尊，如下所示支持多種下游任務(wù)笔横。相對比之前說的Transformer 在圖像中的運(yùn)用（一）VIT(Transformers for Image Recognition at Scale）論文及代碼解讀 之前需要每個像素

一竞滓、 原理

在Transformer種，如果圖像像素太多則我們需要構(gòu)建出更多的特征序列吹缔，這樣就會導(dǎo)致我們的效率降低商佑，所以我們采用了窗口以及分層的形式來替代長序列。

1.1 整體網(wǎng)絡(luò)架構(gòu)

• 得到各Patch特征構(gòu)建的序列（注意這里先卷積得到特征圖厢塘，再對特征圖進(jìn)行切分成Patch）
• 分成計算attention（逐步下采樣過程）
• 其中Block是最核心的茶没， 對attention的計算方法進(jìn)行了改進(jìn)

由下面的圖我們可以看出特征圖大小不斷減小， 但是特征圖的通道數(shù)不斷增加晚碾。

Swin整體網(wǎng)絡(luò)結(jié)構(gòu)

1.1.1 Patch Embedding

下面舉一個例子比如輸入的圖像數(shù)據(jù)為(224, 224, 3)抓半， 輸出(3136, 96)相當(dāng)于序列長度為3136， 每個向量是96維特征格嘁。這里的卷積核我們使用Conv2d(3, 96, kernel_size=(4, 4), stride=(4, 4))笛求。所以3136就是卷積(224 / 4) * (224 / 4)得到的。

這時候我們得到的輸入特征圖為(56, 56, 96)糕簿， 如果默認(rèn)窗口大小為7涣易，所以總共可以分為8 * 8個窗口。則輸出的特征圖為(64(8*8), 7, 7, 96) 之前單位是序列冶伞， 現(xiàn)在單位是窗口(工64個窗口)新症。

1.1.2 Swin Transformer Block

下面我們來看下上面圖中對應(yīng)的Transformer Blocks是什么樣子， 如下圖所示响禽。

Swin Transformer Block

1. W-MSA(計算每個不同窗口自身的注意力機(jī)制(下面不同顏色的矩形代表不同的窗口))

• 3個矩陣
• 64個窗口
• 3個heads
• 7*7的窗口大小(每個窗口有49個token即49個像素)
• 96/3=32個單head特征

所以attention結(jié)果為(64, 3, 49, 49) 每個頭都會得出每個窗口內(nèi)的自注意力（3為頭侯繁，這里可以理解為不同窗口不同頭對應(yīng)窗口的不同token之間的注意力）胖喳。
通過上面的計算我們可以得到新的特征(64, 49, 96), 之后再進(jìn)行reshape操作將其還原到(56, 56, 96)大小特征圖目的就是為了還原輸入特征圖大小(但是其已經(jīng)計算過了attentation), 因為再transformer要經(jīng)過多層輸入大小與輸出大小一般都是相同的。
下面給出了省出來的計算量。

h,w,c分別代表特征圖高度寬度和深度雌贱， M代表窗口大小

矩陣計算評估計算量

這里計算量公式可以參考這篇文章Swin-Transformer網(wǎng)絡(luò)結(jié)構(gòu)詳解啊送。

2. SW-MSA(計算不同窗口之間的注意力機(jī)制)
上面W-MSA是只是知道窗口內(nèi)部的特征，但是我們不知道窗口之間的特征我們可以用SW-MSA機(jī)制來彌補(bǔ)欣孤。這里的主要區(qū)別就是S(shift滑動)馋没，我們?nèi)绾稳プ龌瑒幽?

transformer偏移

https://www.zhihu.com/question/492057377/answer/2213112296

由于不同Windows之間互不重疊泽论，每次進(jìn)行自注意力計算時很顯然就丟失了Windows之間的信息艾少，那么如何在降低計算量的同時保留全局信息呢？Shifted Window應(yīng)運(yùn)而生翼悴。

上面這張圖可以用如下的示意圖理解：

但是還有一個問題原來是4個windows缚够，但是移動之后變成了9個windows，為了能夠做到并行計算應(yīng)該如何解決呢鹦赎？我們可以做如下偏移方法谍椅。

Attention Mask 機(jī)制
因為我們區(qū)域(5,3) (7,1) (8,6,2,0)本來是之間不想連接的，所以我們要單獨(dú)計算各自的區(qū)域的MSA古话。我們借用區(qū)域(5,3)舉例雏吭，這篇博客對于這個解釋非常棒Swin-Transformer網(wǎng)絡(luò)結(jié)構(gòu)詳解， 如下所示：

示例1

示例2

流程1

流程2

流程3

1.1.2 Relative Position Bias

Table 4. Ablation study on the shifted windows approach and different position embedding methods on three benchmarks, using the Swin-T architecture. w/o shifting: all self-attention modules adopt regular window partitioning, without shifting; abs. pos.: absolute position embedding term of ViT; rel. pos.: the default settings with an additional relative position bias term (see Eq. (4)); app.: the first scaled dot-product term in Eq. (4).

上述相對位置小矩陣攤平再拼接就得到下面的大矩陣

偏移從0開始行、列標(biāo)加上M(窗口大小->2*2)-1

行標(biāo)乘上2M-1

行列相加

上述就可以得出我們下面的公式B

1.1.3 PatchMerging

network structure

二、 代碼邏輯解讀

# file: models/swin_transformer.py
# class: SwinTransformer
class SwinTransformer(nn.Module):
    r""" Swin Transformer
        A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows`  -
          https://arxiv.org/pdf/2103.14030

    Args:
        img_size (int | tuple(int)): Input image size. Default 224
        patch_size (int | tuple(int)): Patch size. Default: 4
        in_chans (int): Number of input image channels. Default: 3
        num_classes (int): Number of classes for classification head. Default: 1000
        embed_dim (int): Patch embedding dimension. Default: 96
        depths (tuple(int)): Depth of each Swin Transformer layer.
        num_heads (tuple(int)): Number of attention heads in different layers.
        window_size (int): Window size. Default: 7
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4
        qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float): Override default qk scale of head_dim ** -0.5 if set. Default: None
        drop_rate (float): Dropout rate. Default: 0
        attn_drop_rate (float): Attention dropout rate. Default: 0
        drop_path_rate (float): Stochastic depth rate. Default: 0.1
        norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
        ape (bool): If True, add absolute position embedding to the patch embedding. Default: False
        patch_norm (bool): If True, add normalization after patch embedding. Default: True
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False
    """

    def __init__(self, img_size=224, patch_size=4, in_chans=3, num_classes=1000,
                 embed_dim=96, depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24],
                 window_size=7, mlp_ratio=4., qkv_bias=True, qk_scale=None,
                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1,
                 norm_layer=nn.LayerNorm, ape=False, patch_norm=True,
                 use_checkpoint=False, **kwargs):
        super().__init__()

        self.num_classes = num_classes
        self.num_layers = len(depths)
        self.embed_dim = embed_dim
        self.ape = ape
        self.patch_norm = patch_norm
        self.num_features = int(embed_dim * 2 ** (self.num_layers - 1))
        self.mlp_ratio = mlp_ratio

        # split image into non-overlapping patches
        self.patch_embed = PatchEmbed(
            img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim,
            norm_layer=norm_layer if self.patch_norm else None)
        num_patches = self.patch_embed.num_patches
        patches_resolution = self.patch_embed.patches_resolution
        self.patches_resolution = patches_resolution

        # absolute position embedding
        if self.ape:
            self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
            trunc_normal_(self.absolute_pos_embed, std=.02)

        self.pos_drop = nn.Dropout(p=drop_rate)

        # stochastic depth
        # 這里的drop rate是會隨著模型不同stage不斷提升到我們設(shè)定的rate
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))]  # stochastic depth decay rule

        # build layers
        self.layers = nn.ModuleList()
        for i_layer in range(self.num_layers):
            layer = BasicLayer(dim=int(embed_dim * 2 ** i_layer), # 我們的深度不斷乘上2
                               input_resolution=(patches_resolution[0] // (2 ** i_layer),
                                                 patches_resolution[1] // (2 ** i_layer)),
                               depth=depths[i_layer],
                               num_heads=num_heads[i_layer],
                               window_size=window_size,
                               mlp_ratio=self.mlp_ratio,
                               qkv_bias=qkv_bias, qk_scale=qk_scale,
                               drop=drop_rate, attn_drop=attn_drop_rate,
                               drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
                               norm_layer=norm_layer,
                               downsample=PatchMerging if (i_layer < self.num_layers - 1) else None, # 這里transoformer和patchMerge是連在一起的最后一個沒有transformer只有patchMerge
                               use_checkpoint=use_checkpoint) 
            self.layers.append(layer)

        self.norm = norm_layer(self.num_features)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()

        self.apply(self._init_weights)

2.1 input embedding

# file: models/swin_transformer.py
# class: SwinTransformer
    def forward_features(self, x):
        x = self.patch_embed(x)
        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)  # B L C
        x = self.avgpool(x.transpose(1, 2))  # B C 1
        x = torch.flatten(x, 1)
        return x

這里我們的輸入大小為4(batch), 3(channel), 224(width), 224(height)步绸， 接著進(jìn)入到self.patch_embed操作掺逼。

# file: swin_transformer.py
# class: PatchEmbed
    def forward(self, x):
        B, C, H, W = x.shape
        # FIXME look at relaxing size constraints
        assert H == self.img_size[0] and W == self.img_size[1], \
            f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."
        x = self.proj(x).flatten(2).transpose(1, 2)  # B Ph*Pw C
        if self.norm is not None:
            x = self.norm(x)
        return x

和以往vit一樣吃媒，這里做self.proj就是進(jìn)行卷積操作

# 卷積核大小為4瓤介， stride也是為4， 這樣會導(dǎo)致輸出特征圖為原來的額1/4 -> (56 * 56)
# 輸入輸出channel分別為3和96
self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size

# 這部分flatten操作是將我們的寬度高度展平赘那，輸出shape為(4, 3136(56*56), 96)
x = self.proj(x).flatten(2).transpose(1, 2)

在經(jīng)過self.norm對應(yīng)的操作為nn.LayerNorm刑桑。
接著我們會經(jīng)過我們的self.pos_drop(x)， 這里的self.pos_drop為nn.Dropout(p=drop_rate)操作募舟。
接著進(jìn)行下面各個層的操作（別忘記此時我們的輸入shape為
(4, 3136(56*56), 96)）

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

2.2 Basiclayer

接著上面我們看一下self.layers是如何構(gòu)建的

# file: models/swin_transformer.py
# class: SwinTransformer
        for i_layer in range(self.num_layers):
            layer = BasicLayer(dim=int(embed_dim * 2 ** i_layer),
                               input_resolution=(patches_resolution[0] // (2 ** i_layer),
                                                 patches_resolution[1] // (2 ** i_layer)),
                               depth=depths[i_layer],
                               num_heads=num_heads[i_layer],
                               window_size=window_size,
                               mlp_ratio=self.mlp_ratio,
                               qkv_bias=qkv_bias, qk_scale=qk_scale,
                               drop=drop_rate, attn_drop=attn_drop_rate,
                               drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])],
                               norm_layer=norm_layer,
                               downsample=PatchMerging if (i_layer < self.num_layers - 1) else None,
                               use_checkpoint=use_checkpoint)
            self.layers.append(layer)

# file: models/swin_transformer.py
# class: BasicLayer
class BasicLayer(nn.Module):
    """ A basic Swin Transformer layer for one stage.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resolution.
        depth (int): Number of blocks.
        num_heads (int): Number of attention heads.
        window_size (int): Local window size.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
        norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
        downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
        use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
    """

    def __init__(self, dim, input_resolution, depth, num_heads, window_size,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0.,
                 drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False):

        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.depth = depth
        self.use_checkpoint = use_checkpoint

        # build blocks
        self.blocks = nn.ModuleList([
            SwinTransformerBlock(dim=dim, input_resolution=input_resolution,
                                 num_heads=num_heads, window_size=window_size,
                                 shift_size=0 if (i % 2 == 0) else window_size // 2,
                                 mlp_ratio=mlp_ratio,
                                 qkv_bias=qkv_bias, qk_scale=qk_scale,
                                 drop=drop, attn_drop=attn_drop,
                                 drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
                                 norm_layer=norm_layer)
            for i in range(depth)])

        # patch merging layer
        if downsample is not None:
            self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer)
        else:
            self.downsample = None

    def forward(self, x):
        for blk in self.blocks:
            if self.use_checkpoint:
                x = checkpoint.checkpoint(blk, x)
            else:
                x = blk(x)
        if self.downsample is not None:
            x = self.downsample(x)
        return x

2.2 SwinTransformerBlock

# file: models/swin_transformer.py
# class: SwinTransformerBlock
class SwinTransformerBlock(nn.Module):
    r""" Swin Transformer Block.

    Args:
        dim (int): Number of input channels.
        input_resolution (tuple[int]): Input resulotion.
        num_heads (int): Number of attention heads.
        window_size (int): Window size.
        shift_size (int): Shift size for SW-MSA.
        mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
        qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
        drop (float, optional): Dropout rate. Default: 0.0
        attn_drop (float, optional): Attention dropout rate. Default: 0.0
        drop_path (float, optional): Stochastic depth rate. Default: 0.0
        act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0,
                 mlp_ratio=4., qkv_bias=True, qk_scale=None, drop=0., attn_drop=0., drop_path=0.,
                 act_layer=nn.GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.dim = dim
        self.input_resolution = input_resolution
        self.num_heads = num_heads
        self.window_size = window_size
        self.shift_size = shift_size
        self.mlp_ratio = mlp_ratio
        if min(self.input_resolution) <= self.window_size:
            # if window size is larger than input resolution, we don't partition windows
            self.shift_size = 0
            self.window_size = min(self.input_resolution)
        assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"

        self.norm1 = norm_layer(dim)
        self.attn = WindowAttention(
            dim, window_size=to_2tuple(self.window_size), num_heads=num_heads,
            qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)

        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        self.norm2 = norm_layer(dim)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

        if self.shift_size > 0:
            # calculate attention mask for SW-MSA
            H, W = self.input_resolution
            img_mask = torch.zeros((1, H, W, 1))  # 1 H W 1
            h_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            w_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            cnt = 0
            for h in h_slices:
                for w in w_slices:
                    img_mask[:, h, w, :] = cnt
                    cnt += 1

            mask_windows = window_partition(img_mask, self.window_size)  # nW, window_size, window_size, 1
            mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
            attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
            attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
        else:
            attn_mask = None

        self.register_buffer("attn_mask", attn_mask)

    def forward(self, x):
        H, W = self.input_resolution
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"

        shortcut = x
        x = self.norm1(x)
        x = x.view(B, H, W, C)

        # cyclic shift
        if self.shift_size > 0:
            shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
        else:
            shifted_x = x

        # partition windows
        x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)  # nW*B, window_size*window_size, C

        # W-MSA/SW-MSA
        attn_windows = self.attn(x_windows, mask=self.attn_mask)  # nW*B, window_size*window_size, C

        # merge windows
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
        shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C

        # reverse cyclic shift
        if self.shift_size > 0:
            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
        else:
            x = shifted_x
        x = x.view(B, H * W, C)

        # FFN
        x = shortcut + self.drop_path(x)
        x = x + self.drop_path(self.mlp(self.norm2(x)))

        return x

2.2.1 W-MSA及SW-MSA輸入

我們知道輸入是先經(jīng)過W-MSA再經(jīng)過SW-MSA
經(jīng)過W-MSA是沒有做任何處理的即代碼中shifted_x = x祠斧， 但是對于W-MSA是通過torch.roll的操作進(jìn)行的，代碼如下所示：

shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))

這里有1和2拱礁，分別表示要左右上下移動琢锋，還有就是這里的self.shift_size為負(fù)數(shù)，說明移動完處理之后這里還是要復(fù)原的
如下代碼所示：

        # merge windows
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
        shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C # 第一個block得到（4呢灶， 56吴超， 56， 96）

        # reverse cyclic shift
        if self.shift_size > 0:
            x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
        else:
            x = shifted_x
        x = x.view(B, H * W, C)

最終得到的shape依然是我們原來的輸入(4, 3136, 96)
接著下進(jìn)入如下操作

        # partition windows
        x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)  # nW*B, window_size*window_size, C

比如一開始第一個block我們的得到的第一個輸出shape為(256, 7, 7, 96) 然后我們得到第二個windows為(256, 49, 96)鸯乃。相當(dāng)于256個windows鲸阻， 每個windows由49個像素， 每個像素由96個維度缨睡。
對于上面代碼中的window_partition代碼如下：

def window_partition(x, window_size):
    """
    Args:
        x: (B, H, W, C)
        window_size (int): window size

    Returns:
        windows: (num_windows*B, window_size, window_size, C)
    """
    B, H, W, C = x.shape
    x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
    windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
    return windows

這里的xshape為(4, 8, 7, 8, 7, 96)鸟悴， 我們可以得到windows的數(shù)量為
(H/windows_size) * (W/windows_size) * batch, 這里W,H一開始都為56, windows_size為7, 這里設(shè)置的batch為4, 因此這里我們最終得到的windows shape為(256 7 7 96)。

2.2.2 Attention機(jī)制

上面的輸出之后我們要經(jīng)過我們的Attention機(jī)制奖年。

attn_windows = self.attn(x_windows, mask=self.attn_mask)  # nW*B, window_size*window_size, C

如果x_windows是W-MSA則self.atten_mask為None细诸， 否則會加入atten_mask, 具體代碼如下（詳細(xì)理解可以參考bilibili, 在31分鐘左右 說的非常好）：

        if self.shift_size > 0:
            # calculate attention mask for SW-MSA
            H, W = self.input_resolution
            img_mask = torch.zeros((1, H, W, 1))  # 1 H W 1
            h_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            w_slices = (slice(0, -self.window_size),
                        slice(-self.window_size, -self.shift_size),
                        slice(-self.shift_size, None))
            cnt = 0
            for h in h_slices:
                for w in w_slices:
                    img_mask[:, h, w, :] = cnt
                    cnt += 1

            mask_windows = window_partition(img_mask, self.window_size)  # nW, window_size, window_size, 1
            mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
            attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
            attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
        else:
            attn_mask = None

對應(yīng)上述代碼簡單點(diǎn)就是再不需要做內(nèi)積的地方填入-100, 這樣經(jīng)過softmax的時候就被自動設(shè)置為0了。
下面先給出我們進(jìn)入attention的代碼陋守。

class WindowAttention(nn.Module):
    r""" Window based multi-head self attention (W-MSA) module with relative position bias.
    It supports both of shifted and non-shifted window.

    Args:
        dim (int): Number of input channels.
        window_size (tuple[int]): The height and width of the window.
        num_heads (int): Number of attention heads.
        qkv_bias (bool, optional):  If True, add a learnable bias to query, key, value. Default: True
        qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
        attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
        proj_drop (float, optional): Dropout ratio of output. Default: 0.0
    """

    def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.):

        super().__init__()
        self.dim = dim
        self.window_size = window_size  # Wh, Ww
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5

        # define a parameter table of relative position bias
        self.relative_position_bias_table = nn.Parameter(
            torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))  # 2*Wh-1 * 2*Ww-1, nH

        # get pair-wise relative position index for each token inside the window
        coords_h = torch.arange(self.window_size[0])
        coords_w = torch.arange(self.window_size[1])
        coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
        relative_coords[:, :, 0] += self.window_size[0] - 1  # shift to start from 0
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
        self.register_buffer("relative_position_index", relative_position_index)

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        trunc_normal_(self.relative_position_bias_table, std=.02)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=None):
        """
        Args:
            x: input features with shape of (num_windows*B, N, C)
            mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
        """
        B_, N, C = x.shape
        qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))

        relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
            self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1)  # Wh*Ww,Wh*Ww,nH
        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
        attn = attn + relative_position_bias.unsqueeze(0)

        if mask is not None:
            nW = mask.shape[0]
            attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = self.softmax(attn)
        else:
            attn = self.softmax(attn)

        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

    def extra_repr(self) -> str:
        return f'dim={self.dim}, window_size={self.window_size}, num_heads={self.num_heads}'

    def flops(self, N):
        # calculate flops for 1 window with token length of N
        flops = 0
        # qkv = self.qkv(x)
        flops += N * self.dim * 3 * self.dim
        # attn = (q @ k.transpose(-2, -1))
        flops += self.num_heads * N * (self.dim // self.num_heads) * N
        #  x = (attn @ v)
        flops += self.num_heads * N * N * (self.dim // self.num_heads)
        # x = self.proj(x)
        flops += N * self.dim * self.dim
        return flops

可以看出首先會經(jīng)過self.qkv生成我們的q, k, v矩陣揍堰，內(nèi)部代碼就是很簡單的nn.Linear,

# 這里的`dim`, 我們設(shè)置為96
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
# 這里我們得到的self.qkv shape 為[3, 256, 3, 49, 32] 這里的3分別對應(yīng)qkv， 
# 256個窗口分別做attention嗅义， 
# 剛開始head為3屏歹， 
# 每個窗口有49個元素， 
# 32 代表每個頭有32個維度
# q, k, v shape分別為[256, 3, 49, 32]
q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple) 

接著用得到我們的注意力機(jī)制之碗，如下所示蝙眶，這里的self.scale可以理解為我們的v

        q = q * self.scale
        attn = (q @ k.transpose(-2, -1))

最終讓我們attention與position bias相加， 如下所示獲得我們最終的atten。

attn = attn + relative_position_bias.unsqueeze(0)

這里的position bias 下面解釋幽纷。

2.2.3 Relative Position Bias Table

我們上面說了相對位置偏置矩陣的大小為(2M-1) * (2M-1), 這里的M為windows-size大惺剿（詳細(xì)理解可以參考bilibili, 在56分鐘左右 說的非常好）。

        # define a parameter table of relative position bias
        self.relative_position_bias_table = nn.Parameter(
            torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))  # 2*Wh-1 * 2*Ww-1, nH

        # get pair-wise relative position index for each token inside the window
        coords_h = torch.arange(self.window_size[0])
        coords_w = torch.arange(self.window_size[1])
        coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
        relative_coords[:, :, 0] += self.window_size[0] - 1  # shift to start from 0
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
        self.register_buffer("relative_position_index", relative_position_index)

        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)

        trunc_normal_(self.relative_position_bias_table, std=.02)
        self.softmax = nn.Softmax(dim=-1)

下面就是之前說的經(jīng)softmax友浸， 如果mask不相同索引的我們設(shè)置為-100, 經(jīng)過softmax計算就變成了0.

        if mask is not None:
            nW = mask.shape[0]
            attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
            attn = attn.view(-1, self.num_heads, N, N)
            attn = self.softmax(attn)
        else:
            attn = self.softmax(attn)

        attn = self.attn_drop(attn)

        x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

在經(jīng)過

x = (attn @ v).transpose(1, 2).reshape(B_, N, C)

操作之后我們得到的attention之后的向量為(256, 49. 96)峰尝， self.proj_drop為drop_out。

2.2.4 FFN（殘差操作）

圖中紅色藍(lán)色框的部分做了兩種殘差

最后要做一次殘差連接

        # FFN
        x = shortcut + self.drop_path(x)
        x = x + self.drop_path(self.mlp(self.norm2(x)))

上述說完就完成了我們SwinTransformerBlock的部分了收恢。

3. Patch Merging

Swin整體網(wǎng)絡(luò)結(jié)構(gòu)

class PatchMerging(nn.Module):
    r""" Patch Merging Layer.

    Args:
        input_resolution (tuple[int]): Resolution of input feature.
        dim (int): Number of input channels.
        norm_layer (nn.Module, optional): Normalization layer.  Default: nn.LayerNorm
    """

    def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm):
        super().__init__()
        self.input_resolution = input_resolution
        self.dim = dim
        self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
        self.norm = norm_layer(4 * dim)

    def forward(self, x):
        """
        x: B, H*W, C
        """
        H, W = self.input_resolution
        B, L, C = x.shape
        assert L == H * W, "input feature has wrong size"
        assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even."

        x = x.view(B, H, W, C)

        x0 = x[:, 0::2, 0::2, :]  # B H/2 W/2 C
        x1 = x[:, 1::2, 0::2, :]  # B H/2 W/2 C
        x2 = x[:, 0::2, 1::2, :]  # B H/2 W/2 C
        x3 = x[:, 1::2, 1::2, :]  # B H/2 W/2 C
        x = torch.cat([x0, x1, x2, x3], -1)  # B H/2 W/2 4*C
        x = x.view(B, -1, 4 * C)  # B H/2*W/2 4*C

        x = self.norm(x)
        x = self.reduction(x)

        return x

4. 輸出層

        self.norm = norm_layer(self.num_features)
        self.avgpool = nn.AdaptiveAvgPool1d(1)
        self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()

x = self.norm(x)  # B L C
x = self.avgpool(x.transpose(1, 2))  # B C 1
x = torch.flatten(x, 1)

經(jīng)過平均池化將原來shape由(4, 49, 768)轉(zhuǎn)成(4, 768, 1)后面再接一下全連接層
nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()即可伦意。

參考：
[1] Swin Transformer
[2] 如何看待swin transformer成為ICCV2021的 best paper火窒？
[3] Swin-Transformer網(wǎng)絡(luò)結(jié)構(gòu)詳解