YOLOv8 集成 Swin Transformer V2 修改与训练实战

2. 添加核心模块代码

在 nn/modules/block.py 文件末尾追加 Swin Transformer V2 的相关类定义。这里包含了窗口注意力机制、MLP 以及 CSP 瓶颈层封装。注意，原代码中可能存在重复定义的 WindowAttention 类，清理后保留一份即可，避免命名冲突。

import torch
import torch.nn as nn
import torch.nn.functional as F
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
from .conv import Conv, DWConv, GhostConv, LightConv, RepConv
from .transformer import TransformerBlock
import numpy as np

class WindowAttention(nn.Module):
    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
        self.num_heads = num_heads
        head_dim = dim // num_heads
        self.scale = qk_scale or head_dim ** -0.5
        self.relative_position_bias_table = nn.Parameter(torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads))
        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]))
        coords_flatten = torch.flatten(coords, 1)
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()
        relative_coords[:, :, 0] += self.window_size[0] - 1
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)
        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)
        nn.init.normal_(self.relative_position_bias_table, std=.02)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=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]
        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)
        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()
        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)
        try:
            x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        except:
            x = (attn.half() @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

class WindowAttention_v2(nn.Module):
    def __init__(self, dim, window_size, num_heads, qkv_bias=True, attn_drop=0., proj_drop=0., pretrained_window_size=[0, 0]):
        super().__init__()
        self.dim = dim
        self.window_size = window_size
        self.pretrained_window_size = pretrained_window_size
        self.num_heads = num_heads
        self.logit_scale = nn.Parameter(torch.log(10 * torch.ones((num_heads, 1, 1))), requires_grad=True)
        self.cpb_mlp = nn.Sequential(nn.Linear(2, 512, bias=True), nn.ReLU(inplace=True), nn.Linear(512, num_heads, bias=False))
        relative_coords_h = torch.arange(-(self.window_size[0] - 1), self.window_size[0], dtype=torch.float32)
        relative_coords_w = torch.arange(-(self.window_size[1] - 1), self.window_size[1], dtype=torch.float32)
        relative_coords_table = torch.stack(torch.meshgrid([relative_coords_h, relative_coords_w])).permute(1, 2, 0).contiguous().unsqueeze(0)
        if pretrained_window_size[0] > 0:
            relative_coords_table[:, :, :, 0] /= (pretrained_window_size[0] - 1)
            relative_coords_table[:, :, :, 1] /= (pretrained_window_size[1] - 1)
        else:
            relative_coords_table[:, :, :, 0] /= (self.window_size[0] - 1)
            relative_coords_table[:, :, :, 1] /= (self.window_size[1] - 1)
        relative_coords_table *= 8
        relative_coords_table = torch.sign(relative_coords_table) * torch.log2(torch.abs(relative_coords_table) + 1.0) / np.log2(8)
        self.register_buffer("relative_coords_table", relative_coords_table)
        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]))
        coords_flatten = torch.flatten(coords, 1)
        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]
        relative_coords = relative_coords.permute(1, 2, 0).contiguous()
        relative_coords[:, :, 0] += self.window_size[0] - 1
        relative_coords[:, :, 1] += self.window_size[1] - 1
        relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
        relative_position_index = relative_coords.sum(-1)
        self.register_buffer("relative_position_index", relative_position_index)
        self.qkv = nn.Linear(dim, dim * 3, bias=False)
        if qkv_bias:
            self.q_bias = nn.Parameter(torch.zeros(dim))
            self.v_bias = nn.Parameter(torch.zeros(dim))
        else:
            self.q_bias = None
            self.v_bias = None
        self.attn_drop = nn.Dropout(attn_drop)
        self.proj = nn.Linear(dim, dim)
        self.proj_drop = nn.Dropout(proj_drop)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x, mask=None):
        B_, N, C = x.shape
        qkv_bias = None
        if self.q_bias is not None:
            qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
        qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
        qkv = qkv.reshape(B_, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
        q, k, v = qkv[0], qkv[1], qkv[2]
        attn = (F.normalize(q, dim=-1) @ F.normalize(k, dim=-1).transpose(-2, -1))
        max_tensor = torch.log(torch.tensor(1. / 0.01)).to(self.logit_scale.device)
        logit_scale = torch.clamp(self.logit_scale, max=max_tensor).exp()
        attn = attn * logit_scale
        relative_position_bias_table = self.cpb_mlp(self.relative_coords_table).view(-1, self.num_heads)
        relative_position_bias = 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)
        relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()
        relative_position_bias = 16 * torch.sigmoid(relative_position_bias)
        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)
        try:
            x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
        except:
            x = (attn.half() @ v).transpose(1, 2).reshape(B_, N, C)
        x = self.proj(x)
        x = self.proj_drop(x)
        return x

class Mlp_v2(nn.Module):
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.SiLU, 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 SwinTransformerLayer_v2(nn.Module):
    def __init__(self, dim, num_heads, window_size=7, shift_size=0, mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., drop_path=0., act_layer=nn.SiLU, norm_layer=nn.LayerNorm, pretrained_window_size=0):
        super().__init__()
        self.dim = dim
        self.num_heads = num_heads
        self.window_size = window_size
        self.shift_size = shift_size
        self.mlp_ratio = mlp_ratio
        assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"
        self.norm1 = norm_layer(dim)
        self.attn = WindowAttention_v2(dim, window_size=(self.window_size, self.window_size), num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop, pretrained_window_size=(pretrained_window_size, pretrained_window_size))
        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_v2(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)

    def create_mask(self, H, W):
        img_mask = torch.zeros((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
        def window_partition(x, window_size):
            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
        mask_windows = window_partition(img_mask, self.window_size)
        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))
        return attn_mask

    def forward(self, x):
        _, _, H_, W_ = x.shape
        Padding = False
        if min(H_, W_) < self.window_size or H_ % self.window_size != 0 or W_ % self.window_size != 0:
            Padding = True
            pad_r = (self.window_size - W_ % self.window_size) % self.window_size
            pad_b = (self.window_size - H_ % self.window_size) % self.window_size
            x = F.pad(x, (0, pad_r, 0, pad_b))
        B, C, H, W = x.shape
        L = H * W
        x = x.permute(0, 2, 3, 1).contiguous().view(B, L, C)
        if self.shift_size > 0:
            attn_mask = self.create_mask(H, W).to(x.device)
        else:
            attn_mask = None
        shortcut = x
        x = x.view(B, H, W, C)
        if self.shift_size > 0:
            shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
        else:
            shifted_x = x
        def window_partition(x, window_size):
            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
        def window_reverse(windows, window_size, H, W):
            B = int(windows.shape[0] / (H * W / window_size / window_size))
            x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
            x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
            return x
        x_windows = window_partition(shifted_x, self.window_size)
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)
        attn_windows = self.attn(x_windows, mask=attn_mask)
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
        shifted_x = window_reverse(attn_windows, self.window_size, H, W)
        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)
        x = shortcut + self.drop_path(self.norm1(x))
        x = x + self.drop_path(self.norm2(self.mlp(x)))
        x = x.permute(0, 2, 1).contiguous().view(-1, C, H, W)
        if Padding:
            x = x[:, :, :H_, :W_]
        return x

class SwinTransformer2Block(nn.Module):
    def __init__(self, c1, c2, num_heads, num_layers, window_size=7):
        super().__init__()
        self.conv = None
        if c1 != c2:
            self.conv = Conv(c1, c2)
        self.blocks = nn.Sequential(*[SwinTransformerLayer_v2(dim=c2, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else window_size // 2) for i in range(num_layers)])
    def forward(self, x):
        if self.conv is not None:
            x = self.conv(x)
        x = self.blocks(x)
        return x

class SwinV2_CSPB(nn.Module):
    def __init__(self, c1, c2, n=1, shortcut=False, g=1, e=0.5):
        super(SwinV2_CSPB, self).__init__()
        c_ = int(c2)
        self.cv1 = Conv(c1, c_, 1, 1)
        self.cv2 = Conv(c_, c_, 1, 1)
        self.cv3 = Conv(2 * c_, c2, 1, 1)
        num_heads = c_ // 32
        self.m = SwinTransformer2Block(c_, c_, num_heads, n)
    def forward(self, x):
        x1 = self.cv1(x)
        y1 = self.m(x1)
        y2 = self.cv2(x1)
        return self.cv3(torch.cat((y1, y2), dim=1))

3. 注册新模块

为了让 YOLO 框架识别这些新增的类，需要在 __init__.py 和 task.py 中进行导入和注册。

更新 __init__.py 确保 SwinV2_CSPB 被导出到 __all__ 列表中。

from .block import (
    C1, C2, C3, C3TR, DFL, SPP, SPPF, Bottleneck, BottleneckCSP, C2f, C3Ghost, C3x,
    GhostBottleneck, HGBlock, HGStem, Proto, RepC3, GAM_Attention, ResBlock_CBAM,
    GCT, C3STR, SwinV2_CSPB
)
# ... other imports ...
__all__ = ('Conv', 'Conv2', 'LightConv', 'RepConv', 'DWConv', 'DWConvTranspose2d',
           'ConvTranspose', 'Focus', 'GhostConv', 'ChannelAttention', 'SpatialAttention',
           'CBAM', 'Concat', 'TransformerLayer', 'TransformerBlock', 'MLPBlock',
           'LayerNorm2d', 'DFL', 'HGBlock', 'HGStem', 'SPP', 'SPPF', 'C1', 'C2', 'C3',
           'C2f', 'C3x', 'C3TR', 'C3Ghost', 'GhostBottleneck', 'Bottleneck',
           'BottleneckCSP', 'Proto', 'Detect', 'Segment', 'Pose', 'Classify',
           'TransformerEncoderLayer', 'RepC3', 'RTDETRDecoder', 'AIFI',
           'DeformableTransformerDecoder', 'DeformableTransformerDecoderLayer',
           'MSDeformAttn', 'MLP', 'ResBlock_CBAM', 'CBAM', 'GAM_Attention',
           'GCT', 'C3STR', 'SwinV2_CSPB')

更新 task.py 修正导包语句（注意拼写错误），并在模块检查逻辑中加入新类。

from ultralytics.nn.modules import (
    AIFI, C1, C2, C3, C3TR, SPP, SPPF, Bottleneck, BottleneckCSP, C2f, C3Ghost, C3x,
    Classify, Concat, Conv, Conv2, ConvTranspose, Detect, DWConv, DWConvTranspose2d,
    Focus, GhostBottleneck, GhostConv, HGBlock, HGStem, Pose, RepC3, RepConv,
    RTDETRDecoder, Segment, CBAM, GAM_Attention, ResBlock_CBAM, GCT, C3STR, SwinV2_CSPB
)

# ... inside model definition logic ...
if m in (Classify, Conv, ConvTranspose, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, Focus,
         BottleneckCSP, C1, C2, C2f, C3, C3TR, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, RepC3,
         CBAM, GAM_Attention, ResBlock_CBAM, GCT, C3STR, SwinV2_CSPB):
    # ... existing logic ...

4. 安装与训练

完成上述修改后，务必重新安装本地包以生效。

python setup.py install

随后执行标准的训练命令即可。如果遇到显存溢出，可适当减小 batch size 或调整输入分辨率。Swim Transformer 的计算量较大，请确保硬件资源充足。