Unet

Unet-1

import torch
import torch.nn as nn

def make_divisible(value, divisor=8, min_value=None, min_ratio=0.9):
    if min_value is None:
        min_value = divisor
    new_value = max(min_value, int(value + divisor / 2) // divisor * divisor)
    if new_value < min_ratio * value:
        new_value += divisor
    return new_value

class ConvBlock(nn.Module):
    def __init__(self, in_channels, out_channels):
        super().__init__()
        self.relu = nn.ReLU(inplace=True)
        self.conv_1 = nn.Conv2d(in_channels, out_channels,
                                kernel_size=3, stride=1, padding=1, bias=False)
        self.norm_1 = nn.BatchNorm2d(out_channels)
        self.conv_2 = nn.Conv2d(out_channels, out_channels,
                                kernel_size=3, stride=1, padding=1, bias=False)
        self.norm_2 = nn.BatchNorm2d(out_channels)

    def forward(self, x):
        x = self.relu(self.norm_1(self.conv_1(x)))
        x = self.relu(self.norm_2(self.conv_2(x)))
        return x


class UpSamplingBlock(nn.Module):
    def __init__(self, in_channels, out_channels, up_sampling_mode='transpose'):
        super().__init__()
        if up_sampling_mode == 'transpose':
            self.up_sampling = nn.Sequential(
                nn.ConvTranspose2d(in_channels, out_channels,
                                   kernel_size=2, stride=2, padding=0, bias=False),
                nn.BatchNorm2d(out_channels),
                nn.ReLU(inplace=True)
            )
        else:
            if in_channels == out_channels:
                self.up_sampling = nn.Upsample(scale_factor=2, mode=up_sampling_mode, align_corners=True)
            else:
                self.up_sampling = nn.Sequential(
                    nn.Upsample(scale_factor=2, mode=up_sampling_mode, align_corners=True),
                    nn.Conv2d(in_channels, out_channels,
                              kernel_size=1, stride=1, padding=0, bias=False),
                    nn.BatchNorm2d(out_channels),
                    nn.ReLU(inplace=True)
                )

    def forward(self, x):
        x = self.up_sampling(x)
        return x


class UNet(nn.Module):
    def __init__(self, in_channels=3, classes=10, up_sampling_mode='transpose', channel_ratio=1.0):
        super().__init__()
        self.down_sampling = nn.MaxPool2d(kernel_size=2, stride=2)
        self.left_layer_1 = ConvBlock(in_channels, make_divisible(64 * channel_ratio))
        self.left_layer_2 = ConvBlock(make_divisible(64 * channel_ratio), make_divisible(128 * channel_ratio))
        self.left_layer_3 = ConvBlock(make_divisible(128 * channel_ratio), make_divisible(256 * channel_ratio))
        self.left_layer_4 = ConvBlock(make_divisible(256 * channel_ratio), make_divisible(512 * channel_ratio))

        self.middle_layer = ConvBlock(make_divisible(512 * channel_ratio), make_divisible(1024 * channel_ratio))

        self.up_sampling_1 = UpSamplingBlock(make_divisible(1024 * channel_ratio), make_divisible(512 * channel_ratio),
                                             up_sampling_mode)
        self.right_layer_1 = ConvBlock(make_divisible(1024 * channel_ratio), make_divisible(512 * channel_ratio))
        self.up_sampling_2 = UpSamplingBlock(make_divisible(512 * channel_ratio), make_divisible(256 * channel_ratio),
                                             up_sampling_mode)
        self.right_layer_2 = ConvBlock(make_divisible(512 * channel_ratio), make_divisible(256 * channel_ratio))
        self.up_sampling_3 = UpSamplingBlock(make_divisible(256 * channel_ratio), make_divisible(128 * channel_ratio),
                                             up_sampling_mode)
        self.right_layer_3 = ConvBlock(make_divisible(256 * channel_ratio), make_divisible(128 * channel_ratio))
        self.up_sampling_4 = UpSamplingBlock(make_divisible(128 * channel_ratio), make_divisible(64 * channel_ratio),
                                             up_sampling_mode)
        self.right_layer_4 = ConvBlock(make_divisible(128 * channel_ratio), make_divisible(64 * channel_ratio))

        self.final_layer = nn.Conv2d(make_divisible(64 * channel_ratio), classes,
                                     kernel_size=1, stride=1, padding=0)

    def forward(self, x):
        left_1 = self.left_layer_1(x)
        left_2 = self.left_layer_2(self.down_sampling(left_1))
        left_3 = self.left_layer_3(self.down_sampling(left_2))
        left_4 = self.left_layer_4(self.down_sampling(left_3))

        x = self.middle_layer(self.down_sampling(left_4))

        x = self.up_sampling_1(x)
        x = torch.cat((x, left_4), dim=1)
        x = self.right_layer_1(x)

        x = self.up_sampling_2(x)
        x = torch.cat((x, left_3), dim=1)
        x = self.right_layer_2(x)

        x = self.up_sampling_3(x)
        x = torch.cat((x, left_2), dim=1)
        x = self.right_layer_3(x)

        x = self.up_sampling_4(x)
        x = torch.cat((x, left_1), dim=1)
        x = self.right_layer_4(x)

        x = self.final_layer(x)

        return x


if __name__ == '__main__':
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    in_data = torch.randn(1, 3, 512, 512).to(device)  # b, c, h, w
    model = UNet(in_channels=3, classes= 2, up_sampling_mode='transpose', channel_ratio= 1).to(device)
    out_data = model(in_data)
    print(out_data.shape)

Vit

import torch
import torch.nn as nn
import torch.nn.functional as F

class PatchEmbedding(nn.Module):
    def __init__(self, img_size=224, patch_size=16, in_channels=3, embed_dim=768):
        super().__init__()
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = (img_size // patch_size) ** 2
        self.proj = nn.Conv2d(in_channels, embed_dim, kernel_size=patch_size, stride=patch_size)

    def forward(self, x):
        x = self.proj(x)  # [B, C, H, W] -> [B, embed_dim, n_patches**0.5, n_patches**0.5]
        x = x.flatten(2).transpose(1, 2)  # [B, embed_dim, n_patches] -> [B, n_patches, embed_dim]
        return x

class PositionalEncoding(nn.Module):
    def __init__(self, embed_dim, dropout=0.1, max_len=5000):
        super(PositionalEncoding, self).__init__()
        self.dropout = nn.Dropout(p=dropout)
        pe = torch.zeros(max_len, embed_dim)
        position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, embed_dim, 2).float() * (-torch.log(torch.tensor(10000.0)) / embed_dim))
        pe[:, 0::2] = torch.sin(position * div_term)
        pe[:, 1::2] = torch.cos(position * div_term)
        pe = pe.unsqueeze(0).transpose(0, 1)
        self.register_buffer('pe', pe)

    def forward(self, x):
        x = x + self.pe[:x.size(0), :]
        return self.dropout(x)

class TransformerEncoderLayer(nn.Module):
    def __init__(self, embed_dim, num_heads, dim_feedforward=2048, dropout=0.1):
        super(TransformerEncoderLayer, self).__init__()
        self.self_attn = nn.MultiheadAttention(embed_dim, num_heads, dropout=dropout)
        self.linear1 = nn.Linear(embed_dim, dim_feedforward)
        self.dropout = nn.Dropout(dropout)
        self.linear2 = nn.Linear(dim_feedforward, embed_dim)

        self.norm1 = nn.LayerNorm(embed_dim)
        self.norm2 = nn.LayerNorm(embed_dim)
        self.dropout1 = nn.Dropout(dropout)
        self.dropout2 = nn.Dropout(dropout)

    def forward(self, src):
        src2 = self.norm1(src)
        src2 = self.self_attn(src2, src2, src2)[0]
        src = src + self.dropout1(src2)
        src2 = self.norm2(src)
        src2 = self.linear2(self.dropout(F.relu(self.linear1(src2))))
        src = src + self.dropout2(src2)
        return src

class VisionTransformer(nn.Module):
    def __init__(self, img_size=224, patch_size=16, in_channels=3, embed_dim=768, depth=12,
                 num_heads=12, mlp_ratio=4., num_classes=1000, dropout=0.1):
        super(VisionTransformer, self).__init__()

        self.patch_embedding = PatchEmbedding(img_size, patch_size, in_channels, embed_dim)
        self.pos_encoder = PositionalEncoding(embed_dim, dropout)
        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
        self.encoder_layers = nn.ModuleList([TransformerEncoderLayer(embed_dim, num_heads, int(embed_dim*mlp_ratio), dropout) for _ in range(depth)])
        self.norm = nn.LayerNorm(embed_dim)
        self.fc = nn.Linear(embed_dim, num_classes)

    def forward(self, x):
        B = x.shape[0]
        x = self.patch_embedding(x)  # [B, num_patches, embed_dim]
        cls_tokens = self.cls_token.expand(B, -1, -1)  # [B, 1, embed_dim]
        x = torch.cat((cls_tokens, x), dim=1)  # [B, num_patches+1, embed_dim]
        x = self.pos_encoder(x)  # Add positional encoding

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

        x = self.norm(x)
        x = x[:, 0]  # Use the class token output for classification
        x = self.fc(x)
        return x

if __name__ == '__main__':
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    in_data = torch.randn(2, 3, 224, 224).to(device)  # b, c, h, w
    model = VisionTransformer(img_size= 224, patch_size= 16).to(device)
    out_data = model(in_data)
    print(out_data.shape)

torch.Size([2, 1000])

MAE

import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np

class PatchEmbedding(nn.Module):
    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768):
        super().__init__()
        self.img_size = img_size
        self.patch_size = patch_size
        self.num_patches = (img_size // patch_size) ** 2
        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)

    def forward(self, x):
        x = self.proj(x)  # [B, embed_dim, H/patch_size, W/patch_size]
        x = x.flatten(2).transpose(1, 2)  # [B, num_patches, embed_dim]
        return x

class TransformerEncoder(nn.Module):
    def __init__(self, embed_dim=768, num_heads=12, num_layers=12, mlp_ratio=4.0):
        super().__init__()
        self.layers = nn.ModuleList([
            nn.TransformerEncoderLayer(
                d_model=embed_dim, 
                nhead=num_heads, 
                dim_feedforward=int(embed_dim * mlp_ratio)
            ) 
            for _ in range(num_layers)
        ])
    
    def forward(self, x):
        for layer in self.layers:
            x = layer(x)
        return x

class MAE(nn.Module):
    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768, 
                 encoder_layers=12, decoder_embed_dim=512, decoder_layers=4, mask_ratio=0.75):
        super().__init__()
        self.patch_embed = PatchEmbedding(img_size, patch_size, in_chans, embed_dim)
        self.num_patches = self.patch_embed.num_patches
        self.mask_ratio = mask_ratio

        # Positional encoding
        self.pos_embed = nn.Parameter(torch.zeros(1, self.num_patches, embed_dim))

        # Encoder
        self.encoder = TransformerEncoder(embed_dim, num_heads=12, num_layers=encoder_layers)

        # Mask token
        self.mask_token = nn.Parameter(torch.zeros(1, 1, embed_dim))

        # Decoder
        self.decoder_embed = nn.Linear(embed_dim, decoder_embed_dim)
        self.decoder_pos_embed = nn.Parameter(torch.zeros(1, self.num_patches, decoder_embed_dim))
        self.decoder = TransformerEncoder(decoder_embed_dim, num_heads=8, num_layers=decoder_layers)
        self.decoder_pred = nn.Linear(decoder_embed_dim, patch_size * patch_size * in_chans)  # Pixel values

        # Weight initialization
        self._init_weights()

    def _init_weights(self):
        nn.init.normal_(self.pos_embed, std=0.02)
        nn.init.normal_(self.mask_token, std=0.02)
        nn.init.normal_(self.decoder_pos_embed, std=0.02)

    def random_masking(self, x):
        B, N, D = x.shape
        num_mask = int(N * self.mask_ratio)
        perm = torch.rand(B, N).argsort(dim=1)  # Random shuffle
        mask = perm[:, :num_mask]
        keep = perm[:, num_mask:]
        x_masked = torch.gather(x, dim=1, index=keep.unsqueeze(-1).expand(-1, -1, D))
        return x_masked, mask, keep

    def forward(self, x):
        # Patch embedding
        x = self.patch_embed(x)  # [B, num_patches, embed_dim]
        x += self.pos_embed

        # Random masking
        x_masked, mask, keep = self.random_masking(x)
        print(x_masked.shape, mask.shape, keep.shape)

        # Encoder
        encoded = self.encoder(x_masked)

        # Add mask tokens for decoder input
        B, _, D = x.shape
        mask_tokens = self.mask_token.expand(B, mask.shape[1], D)
        x_full = torch.cat([encoded, mask_tokens], dim=1)
        x_full = torch.gather(x_full, dim=1, index=torch.argsort(torch.cat([keep, mask], dim=1), dim=1).unsqueeze(-1).expand(-1, -1, D))
        # Decoder
        x_full = self.decoder_embed(x_full) + self.decoder_pos_embed
        decoded = self.decoder(x_full)
        pred = self.decoder_pred(decoded)  # [B, num_patches, patch_size*patch_size*in_chans]

        return pred

# Example usage
if __name__ == "__main__":
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    img = torch.randn(2, 3, 224, 224).to(device)
    model = MAE(embed_dim= 768).to(device)
    pred = model(img)  # Output predictions
    print(pred.shape)  # [B, num_patches, patch_size*patch_size*in_chans]

torch.Size([2, 49, 768]) torch.Size([2, 147]) torch.Size([2, 49])
Encoder:torch.Size([2, 49, 768]), mask_tokens:torch.Size([2, 147, 768])
torch.Size([2, 196, 768])

Swin Transformer

代码注释

假设输入图像为：1,3,224,224。仅以stage1（加上PatchMerging）为例

第一步：PatchEmbed：

class PatchEmbed(nn.Module):
    def __init__(self, img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
        ...
        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
        ...
    def forward(self, x):
        ...
        x = self.proj(x)
        x = x.flatten(2).transpose(1, 2)  # B Ph*Pw C
        ...
        return x

参数：输入图片尺寸，分割每个小patch大小，输入通道数量，编码维度 输出：[batch_size, num_pathces, embed_dim] 一般会再补充一个位置编码

...
self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim))
trunc_normal_(self.absolute_pos_embed, std=.02)
...
x = x+ self.absolute_pos_embed

输出：首选卷积处理得到：[1, 96, 56, 56]（224/4=56）然后转化到token形式为：[1, 3136, 96]

第二步：Swin Transformer Block

# 仅以stage1分析
class SwinTransformerBlock(...):
    def __init__(input_resolution, ...):
        ...
    def forward(self, x):
        H, W = self.input_resolution # 56, 56
        B, L, C = x.shape
        ...
        x = x.view(B, H, W, C) # 1,3136,96--> 1, 56, 56, 96

        # 交替转换，切分好不同的window
        if self.shift_size > 0:
            if not self.fused_window_process:
                # 通过 roll 去进行重新排列
                shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
                # partition windows
                x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
            else:
                x_windows = WindowProcess.apply(x, B, H, W, C, -self.shift_size, self.window_size)
        else:
            shifted_x = x
            # partition windows
            x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C

        # 拉平处理
        x_windows = x_windows.view(-1, self.window_size * self.window_size, C)  # nW*B, window_size*window_size, C

        # 计算注意力得分
        attn_windows = self.attn(x_windows, mask=self.attn_mask)  # nW*B, window_size*window_size, C
        # 还原回开始形状
        attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)

再stage1中输入为：[1, 3136, 96]（batch_size=4，相当于对于224改变为56，那么self.input_resolution=(56,56)），不过值得注意的一点是再block中对于W-MSA和SW-MSA是交替进行的，因此会有一个self.shift_size。接下来就是切分不同的window（就在这个window里面计算注意力）（window_size=7）通过window_partition（可以直接理解为：我把图片重新切分成window_size的小图片）处理得到：[64, 7, 7, 96]（1, 56, 56, 96-->64, 7, 7, 96）,然后就是熟悉的拉平处理（就是把图片宽高拉到一起）得到：[64, 49, 96]

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

划分好不同的window之后就直接计算attention-score：

class WindowAttention(...):
    def __init__(self, dim, window_size, num_heads, qkv_bias=True, qk_scale=None, attn_drop=0., proj_drop=0.):
        ...
        # 定义相对位置编码
        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)

    def forward(...):
        # 计算注意力得分，和普通的计算注意力没什么区别
        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))

        # 相对位置编码
        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

计算window-attention score，引入一个 相对位置编码，输入维度是：[64, 49, 96]，然后计算注意力得到：[64, 49, 96]，然后还原为开始形状得到：[64, 7, 7, 96]

# reverse cyclic shift
if self.shift_size > 0:
    if not self.fused_window_process:
        shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
        x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
    else:
        x = WindowProcessReverse.apply(attn_windows, B, H, W, C, self.shift_size, self.window_size)
else:
    shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
    x = shifted_x

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

再经过上述操作之后然后输入到patchmerging，具体操作如下所示：

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

这样一来就实现了尺寸的改变，也就完成了一个完整的satge

值得注意的是： 代码一：

if self.shift_size > 0:
    # calculate attention mask for SW-MSA
    H, W = self.input_resolution # 56 56
    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))
    # 生成h/w的3，3的切片
    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

这部分代码就是只在SW-MSA中进行操作，逐行解读（stage1为例）在处理SW-MSA时需要对特征图进行重新排列，因此就会：

if self.shift_size > 0:
    if not self.fused_window_process:
        # 通过 roll 进行重新排列
        shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
        # partition windows
        x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
    else:
        x_windows = WindowProcess.apply(x, B, H, W, C, -self.shift_size, self.window_size)
    else:
    shifted_x = x
    # partition windows
    x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C

通过上面处理（先进行torch.roll移动一波，这样就可以使最后都有相同的形状，然后再通过window_partition切分一波）保证SW-MSA操作可以实现（都是相同大小）

torch.roll相当于对张量按照shifts=(dima,dimb)方向移动，负数：向左/上移动，正数：向右/下移动

# --------------------------------------------------------
# Swin Transformer
# Copyright (c) 2021 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ze Liu
# --------------------------------------------------------

import torch
import torch.nn as nn
import torch.utils.checkpoint as checkpoint
from timm.models.layers import DropPath, to_2tuple, trunc_normal_

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


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


def window_reverse(windows, window_size, H, W):
    """
    Args:
        windows: (num_windows*B, window_size, window_size, C)
        window_size (int): Window size
        H (int): Height of image
        W (int): Width of image

    Returns:
        x: (B, H, W, C)
    """
    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


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


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
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. Default: False
    """

    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,
                 fused_window_process=False):
        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)
        self.fused_window_process = fused_window_process

    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:
            if not self.fused_window_process:
                shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
                # partition windows
                x_windows = window_partition(shifted_x, self.window_size)  # nW*B, window_size, window_size, C
            else:
                x_windows = WindowProcess.apply(x, B, H, W, C, -self.shift_size, self.window_size)
        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)

        # reverse cyclic shift
        if self.shift_size > 0:
            if not self.fused_window_process:
                shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
                x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
            else:
                x = WindowProcessReverse.apply(attn_windows, B, H, W, C, self.shift_size, self.window_size)
        else:
            shifted_x = window_reverse(attn_windows, self.window_size, H, W)  # B H' W' C
            x = shifted_x
        x = x.view(B, H * W, C)
        x = shortcut + self.drop_path(x)

        # FFN
        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


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

    def extra_repr(self) -> str:
        return f"input_resolution={self.input_resolution}, dim={self.dim}"

    def flops(self):
        H, W = self.input_resolution
        flops = H * W * self.dim
        flops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim
        return flops


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.
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. 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,
                 fused_window_process=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,
                                 fused_window_process=fused_window_process)
            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

    def extra_repr(self) -> str:
        return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}"

    def flops(self):
        flops = 0
        for blk in self.blocks:
            flops += blk.flops()
        if self.downsample is not None:
            flops += self.downsample.flops()
        return flops


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


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
        fused_window_process (bool, optional): If True, use one kernel to fused window shift & window partition for acceleration, similar for the reversed part. 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, fused_window_process=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
        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),
                               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,
                               fused_window_process=fused_window_process)
            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)

    def _init_weights(self, m):
        if isinstance(m, nn.Linear):
            trunc_normal_(m.weight, std=.02)
            if isinstance(m, nn.Linear) and m.bias is not None:
                nn.init.constant_(m.bias, 0)
        elif isinstance(m, nn.LayerNorm):
            nn.init.constant_(m.bias, 0)
            nn.init.constant_(m.weight, 1.0)

    @torch.jit.ignore
    def no_weight_decay(self):
        return {'absolute_pos_embed'}

    @torch.jit.ignore
    def no_weight_decay_keywords(self):
        return {'relative_position_bias_table'}

    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

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

    def flops(self):
        flops = 0
        flops += self.patch_embed.flops()
        for i, layer in enumerate(self.layers):
            flops += layer.flops()
        flops += self.num_features * self.patches_resolution[0] * self.patches_resolution[1] // (2 ** self.num_layers)
        flops += self.num_features * self.num_classes
        return flops

if __name__ == '__main__':
    x = torch.randn(1, 3, 224, 224)
    model = SwinTransformer()
    out = model(x)
    print(out.shape)

torch.Size([1, 1000])