【AI】Swin Transformer源码解析

0. Swin Transformer简介

Swin Transformer指出，CV中的Token(处理单元)的大小不固定，并且数量相较于NLP要多。为解决这两个问题，Swin Transformer不仅使用了分层结构（金字塔结构），同时还提出了一种线性复杂度的Attention计算，最终得到了几种不同大小的结构。

看图说话：
（1）使用了类似卷积神经网络中的层次化构建方法，通过金字塔分层结构和下采样拓宽了感受野，这样的backbone有助于在此基础上构建目标检测，实例分割等任务
（2）在Swin Transformer中使用了Windows Multi-Head Self-Attention(W-MSA)的概念，将特征图划分成了多个不相交的区域（Window），Multi-Head Self-Attention只在每个窗口（Window）内进行。这样的结构可以减少计算量，但也会导致不同window中的结果缺少关联性，所以又添加了滑动窗口的Windows Multi-Head Self-Attention(W-MSA)

所以Swin Transform 主要的要点就是Patch Merging, W-MSA, SW-MSA, 相对位置编码这四个方面。

Source code:

https://github.com/microsoft/Swin-Transformer

1.源码解析

由Swin Tranformer的网络结构图我们可以看出来数据处理流程：Patch Embeding -> Basic Layer (SwinTransBlock * n -> Patch Merging) -> avgpool -> mlp head

PatchEmbedding

一张图展示数据处理的方式：

class PatchEmbed(nn.Module):
    r""" Image to Patch Embedding

    Args:
        img_size (int): Image size.  Default: 224.
        patch_size (int): Patch token size. Default: 4.
        in_chans (int): Number of input image channels. Default: 3.
        embed_dim (int): Number of linear projection output channels. Default: 96.
        norm_layer (nn.Module, optional): Normalization layer. Default: None
    """

    def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
        super().__init__()
        img_size = to_2tuple(img_size)
        patch_size = to_2tuple(patch_size)
        patches_resolution = [img_size[0] // patch_size[0], img_size[1] // patch_size[1]]
        self.img_size = img_size
        self.patch_size = patch_size
        self.patches_resolution = patches_resolution
        self.num_patches = patches_resolution[0] * patches_resolution[1]

        self.in_chans = in_chans
        self.embed_dim = embed_dim

        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        if norm_layer is not None:
            self.norm = norm_layer(embed_dim)
        else:
            self.norm = None

    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

    def flops(self):
        Ho, Wo = self.patches_resolution
        flops = Ho * Wo * self.embed_dim * self.in_chans * (self.patch_size[0] * self.patch_size[1])
        if self.norm is not None:
            flops += Ho * Wo * self.embed_dim
        return flops

Swin Transformer Block

这里的LN和MLP层都是比较熟悉的处理方式，不同于VIT的是，这里将Multi-head Attention换成了W-MSA和SW-MSA两种结构，这里的W就是Swin中的window，SW就是shift-window，是我们需要着重了解的。

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
        #首先确定shift_size是多少，如果输入的分辨率小于窗口尺寸，wmsa和msa等价，因此shift设为0，窗口调整为input大小。
        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
            #slice 截取固定数量的元素
            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))
            #给每个window编号
            cnt = 0
            for h in h_slices:
                for w in w_slices:
                    img_mask[:, h, w, :] = cnt
                    cnt += 1
            #mask window也进window切分部分，执行roll等操作。
            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)
            #通过广播操作相减，使得index相同的部分attn_mask上为0，后续根据attn_mask上的值处理
            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

        #定义一个不可学习的变量，attn_mask为属性名。
        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 这里使用shift_size就是区分W-MSA和SW-MSA的结构
        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

    def extra_repr(self) -> str:
        return f"dim={self.dim}, input_resolution={self.input_resolution}, num_heads={self.num_heads}, " \
               f"window_size={self.window_size}, shift_size={self.shift_size}, mlp_ratio={self.mlp_ratio}"

    def flops(self):
        flops = 0
        H, W = self.input_resolution
        # norm1
        flops += self.dim * H * W
        # W-MSA/SW-MSA
        nW = H * W / self.window_size / self.window_size
        flops += nW * self.attn.flops(self.window_size * self.window_size)
        # mlp
        flops += 2 * H * W * self.dim * self.dim * self.mlp_ratio
        # norm2
        flops += self.dim * H * W
        return flops

SW-MSA处理方式
可以通过roll的方式将切碎的小window从左上角挪动到有右下角，拼一起，计算时候把不同window间的值mask掉，计算完成后再roll回来。

这里的操作源码写的比较精炼,这里的shift是负数，但roll回来的时候，就变为正值了

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

相对位置编码实现

coords_h = torch.arange(self.window_size[0])
coords_w = torch.arange(self.window_size[1])
#meshgrid，生成网格，沿着x，y轴的方向复制coords_h和coords_w
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)

mask的实现形式

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
    #slice 截取固定数量的元素
    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))
    #给每个window编号
    cnt = 0
    for h in h_slices:
        for w in w_slices:
            img_mask[:, h, w, :] = cnt
            cnt += 1
    #mask window也进window切分部分，执行roll等操作。
    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)
    #通过广播操作相减，使得index相同的部分attn_mask上为0，后续根据attn_mask上的值处理
    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

参考文章：
Swin-Transformer网络结构详解