VIT代码解析 （bubbliiiing版本）

1.代码流程

以VIT-B/16为例，模型结构下图所示

将以下列顺序进行解析

1.PatchEmbed

2.Class token

3.Position Embeding

4.Dropout

5.Encoder Block

6.MLP block

7.Layer Norm

8.Extract Class Token

9.MLP Head

2.代码块

2.1?PatchEmbed

对应网络结构图的这一块，输入为224*224*3的图片，经过Conv2d（k=16,s=16）得到768个大小为14*14的分割块，再经过展平和维度变换得到768*196，即768个序列长度为196的图像特征。

class PatchEmbed(nn.Module):
    def __init__(self, input_shape=[224, 224], patch_size=16, in_chans=3, num_features=768, norm_layer=None, flatten=True):
        super().__init__()
        self.num_patches    = (input_shape[0] // patch_size) * (input_shape[1] // patch_size)
        self.flatten        = flatten

        self.proj = nn.Conv2d(in_chans, num_features, kernel_size=patch_size, stride=patch_size)
        self.norm = norm_layer(num_features) if norm_layer else nn.Identity()

    def forward(self, x):
        x = self.proj(x)
        if self.flatten:
            x = x.flatten(2).transpose(1, 2)  # BCHW -> BNC
        x = self.norm(x)
        return x

2.2 Class token

对应这一块

classtoken部分是transformer的分类特征。用于堆叠到序列化后的图片特征中，作为一个单位的序列特征进行特征提取。? ?在利用步长为16x16的卷积将输入图片划分成14x14的部分后，将14x14部分的特征平铺，一幅图片会存在序列长度为196的特征。

此时生成一个classtoken，将classtoken堆叠到序列长度为196的特征上，获得一个序列长度为197的特征。在特征提取的过程中，classtoken会与图片特征进行特征的交互。最终分类时，我们取出classtoken的特征，利用全连接分类。

num_features就是patchembed的输出结果

self.cls_token      = nn.Parameter(torch.zeros(1, 1, num_features))

2.3Position Embeding?

对应这一块

为网络提取到的特征添加上位置信息。
以输入图片为224, 224, 3为例，我们获得的序列化后的图片特征为196, 768。加上classtoken后就是197, 768
此时生成的pos_Embedding的shape也为197, 768，代表每一个特征的位置信息。
?

       self.pos_embed      = nn.Parameter(torch.zeros(1, num_patches + 1, num_features))

?2.4 Dropout

对应这一块

def drop_path(x, drop_prob: float = 0., training: bool = False):
    if drop_prob == 0. or not training:
        return x
    keep_prob       = 1 - drop_prob
    shape           = (x.shape[0],) + (1,) * (x.ndim - 1)
    random_tensor   = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device)
    random_tensor.floor_() 
    output          = x.div(keep_prob) * random_tensor
    return output

class DropPath(nn.Module):
    def __init__(self, drop_prob=None):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training)

?2.5 Encoder Block

对应这一块

Encoder Block的作用是将输入序列进行编码，生成对应的语义向量。它由多个子层组成，包括自注意力机制（Self-Attention）和前馈神经网络（Feed-Forward Neural Network）。自注意力机制用于捕捉输入序列中不同位置之间的依赖关系，而前馈神经网络则用于对每个位置的特征进行非线性变换。通过多个Encoder Block的堆叠，可以逐步提取输入序列的语义信息，并生成对应的语义向量c，用于后续的解码工作。?

class Block(nn.Module):
    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0.,
                 drop_path=0., act_layer=GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.norm1      = norm_layer(dim)
        self.attn       = Attention(dim, num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop)
        self.norm2      = norm_layer(dim)
        self.mlp        = Mlp(in_features=dim, hidden_features=int(dim * mlp_ratio), act_layer=act_layer, drop=drop)
        self.drop_path  = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        
    def forward(self, x):
        x = x + self.drop_path(self.attn(self.norm1(x)))
        x = x + self.drop_path(self.mlp(self.norm2(x)))
        return x

2.5.1Multi-Head Attention

对应这一块

# ? Attention机制
# ? 将输入的特征qkv特征进行划分，首先生成query, key, value。query是查询向量、key是键向量、v是值向量。
# ? 然后利用 查询向量query 点乘 转置后的键向量key，这一步可以通俗的理解为，利用查询向量去查询序列的特征，获得序列每个部分的重要程度score。
# ? 然后利用 score 点乘 value，这一步可以通俗的理解为，将序列每个部分的重要程度重新施加到序列的值上去。?

class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., proj_drop=0.):
        super().__init__()
        self.num_heads  = num_heads
        self.scale      = (dim // num_heads) ** -0.5

        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)

    def forward(self, x):
        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]

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

2.6 MLP block

对应这一块

MLP在ViT（Vision Transformer）中的作用是替代self-attention机制，用于实现特征的混合和整合。在ViT中，首先将输入图像分割成一系列的小块，然后通过patch embedding将每个小块映射为一个向量表示。接下来，这些向量会经过多个Mixer Layer进行处理。

每个Mixer Layer由两个部分组成：channel-mixing MLPs和token-mixing MLPs。channel-mixing MLPs用于在每个小块的通道维度上进行特征混合，以捕捉不同通道之间的关系。而token-mixing MLPs则用于在每个小块的位置维度上进行特征混合，以捕捉不同位置之间的关系。

与传统的self-attention机制相比，MLP在ViT中的作用是通过全连接层来建模特征之间的关系，而不是通过注意力权重来计算特征的重要性。这种替代方式使得ViT在处理图像任务时能够取得和传统的self-attention机制相媲美的结果。

因为token-mixing MLPs对输入tokens的顺序非常敏感，所以在Mixer中不适用positional encoding。这意味着Mixer不会考虑输入图像的位置信息，而仅仅关注特征之间的关系。这种设计选择使得Mixer在处理图像时更加高效，并且能够取得很好的性能。

?

class Mlp(nn.Module):
    """ MLP as used in Vision Transformer, MLP-Mixer and related networks
    """
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=GELU, drop=0.):
        super().__init__()
        out_features    = out_features or in_features
        hidden_features = hidden_features or in_features
        drop_probs      = (drop, drop)

        self.fc1    = nn.Linear(in_features, hidden_features)
        self.act    = act_layer()
        self.drop1  = nn.Dropout(drop_probs[0])
        self.fc2    = nn.Linear(hidden_features, out_features)
        self.drop2  = nn.Dropout(drop_probs[1])

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop1(x)
        x = self.fc2(x)
        x = self.drop2(x)
        return x

2.6.1GELU

对应这一块

class GELU(nn.Module):
    def __init__(self):
        super(GELU, self).__init__()

    def forward(self, x):
        return 0.5 * x * (1 + F.tanh(np.sqrt(2 / np.pi) * (x + 0.044715 * torch.pow(x,3))))

2.7 Layer Norm

对应这一块

       self.norm = norm_layer(num_features)

?

2.8 Extract Class Token

对应这一块

        return x[:, 0]

?

2.9MLP Head

对应这一块

       self.head = nn.Linear(num_features, num_classes) if num_classes > 0 else nn.Identity()

3.完整代码

from functools import partial

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

#--------------------------------------#
#   Gelu激活函数的实现
#   利用近似的数学公式
#--------------------------------------#
class GELU(nn.Module):
    def __init__(self):
        super(GELU, self).__init__()

    def forward(self, x):
        return 0.5 * x * (1 + F.tanh(np.sqrt(2 / np.pi) * (x + 0.044715 * torch.pow(x,3))))

def drop_path(x, drop_prob: float = 0., training: bool = False):
    if drop_prob == 0. or not training:
        return x
    keep_prob       = 1 - drop_prob
    shape           = (x.shape[0],) + (1,) * (x.ndim - 1)
    random_tensor   = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device)
    random_tensor.floor_() 
    output          = x.div(keep_prob) * random_tensor
    return output

class DropPath(nn.Module):
    def __init__(self, drop_prob=None):
        super(DropPath, self).__init__()
        self.drop_prob = drop_prob

    def forward(self, x):
        return drop_path(x, self.drop_prob, self.training)

class PatchEmbed(nn.Module):
    def __init__(self, input_shape=[224, 224], patch_size=16, in_chans=3, num_features=768, norm_layer=None, flatten=True):
        super().__init__()
        self.num_patches    = (input_shape[0] // patch_size) * (input_shape[1] // patch_size)
        self.flatten        = flatten

        self.proj = nn.Conv2d(in_chans, num_features, kernel_size=patch_size, stride=patch_size)
        self.norm = norm_layer(num_features) if norm_layer else nn.Identity()

    def forward(self, x):
        x = self.proj(x)
        if self.flatten:
            x = x.flatten(2).transpose(1, 2)  # BCHW -> BNC
        x = self.norm(x)
        return x

#--------------------------------------------------------------------------------------------------------------------#
#   Attention机制
#   将输入的特征qkv特征进行划分，首先生成query, key, value。query是查询向量、key是键向量、v是值向量。
#   然后利用 查询向量query 点乘 转置后的键向量key，这一步可以通俗的理解为，利用查询向量去查询序列的特征，获得序列每个部分的重要程度score。
#   然后利用 score 点乘 value，这一步可以通俗的理解为，将序列每个部分的重要程度重新施加到序列的值上去。
#--------------------------------------------------------------------------------------------------------------------#
class Attention(nn.Module):
    def __init__(self, dim, num_heads=8, qkv_bias=False, attn_drop=0., proj_drop=0.):
        super().__init__()
        self.num_heads  = num_heads
        self.scale      = (dim // num_heads) ** -0.5

        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)

    def forward(self, x):
        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]

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

class Mlp(nn.Module):
    """ MLP as used in Vision Transformer, MLP-Mixer and related networks
    """
    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=GELU, drop=0.):
        super().__init__()
        out_features    = out_features or in_features
        hidden_features = hidden_features or in_features
        drop_probs      = (drop, drop)

        self.fc1    = nn.Linear(in_features, hidden_features)
        self.act    = act_layer()
        self.drop1  = nn.Dropout(drop_probs[0])
        self.fc2    = nn.Linear(hidden_features, out_features)
        self.drop2  = nn.Dropout(drop_probs[1])

    def forward(self, x):
        x = self.fc1(x)
        x = self.act(x)
        x = self.drop1(x)
        x = self.fc2(x)
        x = self.drop2(x)
        return x

class Block(nn.Module):
    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, drop=0., attn_drop=0.,
                 drop_path=0., act_layer=GELU, norm_layer=nn.LayerNorm):
        super().__init__()
        self.norm1      = norm_layer(dim)
        self.attn       = Attention(dim, num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop)
        self.norm2      = norm_layer(dim)
        self.mlp        = Mlp(in_features=dim, hidden_features=int(dim * mlp_ratio), act_layer=act_layer, drop=drop)
        self.drop_path  = DropPath(drop_path) if drop_path > 0. else nn.Identity()
        
    def forward(self, x):
        x = x + self.drop_path(self.attn(self.norm1(x)))
        x = x + self.drop_path(self.mlp(self.norm2(x)))
        return x
        
class VisionTransformer(nn.Module):
    def __init__(
            self, input_shape=[224, 224], patch_size=16, in_chans=3, num_classes=1000, num_features=768,
            depth=12, num_heads=12, mlp_ratio=4., qkv_bias=True, drop_rate=0.1, attn_drop_rate=0.1, drop_path_rate=0.1,
            norm_layer=partial(nn.LayerNorm, eps=1e-6), act_layer=GELU
        ):
        super().__init__()
        #-----------------------------------------------#
        #   224, 224, 3 -> 196, 768
        #-----------------------------------------------#
        self.patch_embed    = PatchEmbed(input_shape=input_shape, patch_size=patch_size, in_chans=in_chans, num_features=num_features)
        num_patches         = (224 // patch_size) * (224 // patch_size)
        self.num_features   = num_features
        self.new_feature_shape = [int(input_shape[0] // patch_size), int(input_shape[1] // patch_size)]
        self.old_feature_shape = [int(224 // patch_size), int(224 // patch_size)]

        #--------------------------------------------------------------------------------------------------------------------#
        #   classtoken部分是transformer的分类特征。用于堆叠到序列化后的图片特征中，作为一个单位的序列特征进行特征提取。
        #
        #   在利用步长为16x16的卷积将输入图片划分成14x14的部分后，将14x14部分的特征平铺，一幅图片会存在序列长度为196的特征。
        #   此时生成一个classtoken，将classtoken堆叠到序列长度为196的特征上，获得一个序列长度为197的特征。
        #   在特征提取的过程中，classtoken会与图片特征进行特征的交互。最终分类时，我们取出classtoken的特征，利用全连接分类。
        #--------------------------------------------------------------------------------------------------------------------#
        #   196, 768 -> 197, 768
        self.cls_token      = nn.Parameter(torch.zeros(1, 1, num_features))
        #--------------------------------------------------------------------------------------------------------------------#
        #   为网络提取到的特征添加上位置信息。
        #   以输入图片为224, 224, 3为例，我们获得的序列化后的图片特征为196, 768。加上classtoken后就是197, 768
        #   此时生成的pos_Embedding的shape也为197, 768，代表每一个特征的位置信息。
        #--------------------------------------------------------------------------------------------------------------------#
        #   197, 768 -> 197, 768
        self.pos_embed      = nn.Parameter(torch.zeros(1, num_patches + 1, num_features))
        self.pos_drop       = nn.Dropout(p=drop_rate)

        #-----------------------------------------------#
        #   197, 768 -> 197, 768  12次
        #-----------------------------------------------#
        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]
        self.blocks = nn.Sequential(
            *[
                Block(
                    dim         = num_features, 
                    num_heads   = num_heads, 
                    mlp_ratio   = mlp_ratio, 
                    qkv_bias    = qkv_bias, 
                    drop        = drop_rate,
                    attn_drop   = attn_drop_rate, 
                    drop_path   = dpr[i], 
                    norm_layer  = norm_layer, 
                    act_layer   = act_layer
                )for i in range(depth)
            ]
        )
        self.norm = norm_layer(num_features)
        self.head = nn.Linear(num_features, num_classes) if num_classes > 0 else nn.Identity()

    def forward_features(self, x):
        x           = self.patch_embed(x)
        cls_token   = self.cls_token.expand(x.shape[0], -1, -1) 
        x           = torch.cat((cls_token, x), dim=1)
        
        cls_token_pe = self.pos_embed[:, 0:1, :]
        img_token_pe = self.pos_embed[:, 1: , :]

        img_token_pe = img_token_pe.view(1, *self.old_feature_shape, -1).permute(0, 3, 1, 2)
        img_token_pe = F.interpolate(img_token_pe, size=self.new_feature_shape, mode='bicubic', align_corners=False)
        img_token_pe = img_token_pe.permute(0, 2, 3, 1).flatten(1, 2)
        pos_embed = torch.cat([cls_token_pe, img_token_pe], dim=1)

        x = self.pos_drop(x + pos_embed)
        x = self.blocks(x)
        x = self.norm(x)
        return x[:, 0]

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

    def freeze_backbone(self):
        backbone = [self.patch_embed, self.cls_token, self.pos_embed, self.pos_drop, self.blocks[:8]]
        for module in backbone:
            try:
                for param in module.parameters():
                    param.requires_grad = False
            except:
                module.requires_grad = False

    def Unfreeze_backbone(self):
        backbone = [self.patch_embed, self.cls_token, self.pos_embed, self.pos_drop, self.blocks[:8]]
        for module in backbone:
            try:
                for param in module.parameters():
                    param.requires_grad = True
            except:
                module.requires_grad = True

    
def vit_b_16(input_shape=[224, 224], pretrained=False, num_classes=1000):
    model = VisionTransformer(input_shape)
    if pretrained:
        model.load_state_dict(torch.load("model_data/vit-patch_16.pth"))

    if num_classes!=1000:
        model.head = nn.Linear(model.num_features, num_classes)
    return model