Transformer

一、Transformer

1、简介

创新、模型效果
通用的模块注意力机制
应用领域：cv nlp 信号处理
视觉、文本、语音、信号
核心：提特征的方法提的更好

应用NLP的文本任务

nlp word2vec 词向量
每个词都是一个向量
不同的语境中一个词的含义不同

2、Attention 注意力机制

权重控制

语言：感兴趣的
图像：指定需要关注

（1）self-attention

作用：同样的词，语境不同，含义不同

基于权重项，重构特征
重构：计算与其他词之间的关系，对每个向量做重构
输入向量——重构——输出重构完的向量

attention帮助提取特征，包括局部特征

self-attention与attention的区别：

self-attention：我自己的词，和自己上下文进行计算
attention：与其他词进行计算

（2）Transformer细节

Input Embedding Queries Keys Values

以NLP中为例：

x1、x2为embeding得到的结果
由x1与x1、x2之间的关系
x1 询问——Queries q1
问自己——回答k1——q1k1算内积
x2 别人问我，给的答案 key2——q1k2
每个词都会问其他词与自己之间的关系

x2问x1——q2k1

内积越大，关系越紧密，值越大；垂直内积为0

初始化权重参数矩阵，分别进行迭代优化，最终输出最好的值
q、k、v是由训练得到——通过权重获得

q 要去查询
k 等着被查
v 实际的特征信息

数值越大，特征越重要
dk向量维度——排除掉维度对结果的影响
softmax归一化 0~1

（3）multi-header 多头注意力机制

每个词与其他词的关系由模型来定
同一个词，不同模型，关系结果不同
特征拼接
很多个特征，所有特征拼接在一起
再通过全连接层——降维

x1 q11 k11 q21 k21
x2 q12 k12 q22 k22

问题：先算后算q1k2，都无所谓
解决：需要加上序号

位置信息表达：对每个词加上单独的表达，位置编码

3、图像任务Transformer

NLp最火论文: Attention is all your need

视觉中的Attention：只关注主体

multi-head self attention 多头自注意力机制--类似于Group Convolution

将值拆分
concat连接

文本——分词——每个词——上下文间关系
图像——分块embedding(固定大小分块)——确定区域——按顺序排列组合
第一块与第一块到 N块的关系，最后叠加
Transformer——特征提取器

第一次卷积，卷积核区域很小，非常小特征
第二次，在前面的小块基础上
获取的特征区域逐渐变大，慢慢变全局
深度——感受野非常大
CNN——感受野
CNN问题：要获得全局视野需要很多层
Transformer与CNN相比的优势：

第一层：即考虑自己，又考虑全局
不需要堆叠，直接可以获得全局信息
Transformer需要训练数据到位（预训练模型、训练改进）

二、ViT网络

1、Vision Transformer(ViT) 2020 CVPR

Transformer动图：

ViT/16流程

16个patchEmbedding层(Linear Projection of Flattened Patches)
Token(向量，加上posiiton enbedding)
Transformer Encoder(重复堆叠L次)
MLP Head
得到分类结果

Linear Projection of Flattened Patches(Embedding层)
Transformer Encoder
MLP Head(最终用于分类的层结构)

Embedding层：要求输入token(向量)序列，二维矩阵[num_token, token_dim]

2、实现步骤

通过一个卷积层来实现，以ViTB/16为例

使用卷积核为16×16，stride为16，卷积核个数为768 [224, 224, 3] → [14, 14, 768] → 展平 → [198, 768]
需要加上[class]token及Position Embedding，都是可训练参数
- 拼接[class]token: Cat{[1, 768], [196, 768]} → [197, 768]
- 叠加Position Embedding：[197, 768] → [197, 768]
- 位置编码相似度——余弦相似度
Layer：Transformer Encoder中重复堆叠Encoder Block的次数
Hidden Size：Embedding层后每个token的dim（向量的长度）
MLP Size：Transformer Encoder中MLP Block第一个全连接的节点个数（是Hidden Size的四倍）
Heads：Transformer中Multi-Head Attention的heads数
- Params：参数个数
Hybrid：传统卷积网络提取提取特征 ResNet StdConv2d;所有BatchNorm层替换成GroupNorm层；stage4中3个block移至stage3中

网络参数

Model	Patch Size	Layers	Hidden Size D	MLP size	Heads	Params
ViT-Base	16×16	12	768	3072	12	86M
Vit-Large	16×16	24	1024	4096	16	307M
ViT-Huge	14×14	32	1280	5120	16	632M

3、混合模型——R50-ViTB16

4、网络代码

"""
original code from rwightman:
https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py
"""
from functools import partial
from collections import OrderedDictimport torch
import torch.nn as nn# 使用时，需要下载.pth预训练模型，先学习
# 因为需要在非常大的训练集中训练，才会有很好的效果def drop_path(x, drop_prob: float = 0., training: bool = False):"""Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).This is the same as the DropConnect impl I created for EfficientNet, etc networks, however,the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper...See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted forchanging the layer and argument names to 'drop path' rather than mix DropConnect as a layer name and use'survival rate' as the argument."""if drop_prob == 0. or not training:return xkeep_prob = 1 - drop_probshape = (x.shape[0],) + (1,) * (x.ndim - 1)  # work with diff dim tensors, not just 2D ConvNetsrandom_tensor = keep_prob + torch.rand(shape, dtype=x.dtype, device=x.device)random_tensor.floor_()  # binarizeoutput = x.div(keep_prob) * random_tensorreturn outputclass DropPath(nn.Module):"""Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks)."""def __init__(self, drop_prob=None):super(DropPath, self).__init__()self.drop_prob = drop_probdef forward(self, x):return drop_path(x, self.drop_prob, self.training)# Patch Embedding
class PatchEmbed(nn.Module):"""2D Image to Patch Embedding"""def __init__(self, img_size=224, patch_size=16, in_c=3, embed_dim=768, norm_layer=None):super().__init__()img_size = (img_size, img_size)patch_size = (patch_size, patch_size)self.img_size = img_sizeself.patch_size = patch_sizeself.grid_size = (img_size[0] // patch_size[0], img_size[1] // patch_size[1])  # (14, 14)self.num_patches = self.grid_size[0] * self.grid_size[1]  # 14×14=196self.proj = nn.Conv2d(in_c, embed_dim, kernel_size=patch_size, stride=patch_size)# nn.Identity() 建立一个输入层，什么都不做self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()def forward(self, x):B, C, H, W = x.shapeassert 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]})."# flatten: [B, C, H, W] -> [B, C, HW]# transpose: [B, C, HW] -> [B, HW, C]x = self.proj(x).flatten(2).transpose(1, 2)x = self.norm(x)return xclass Attention(nn.Module):def __init__(self,dim,  # 输入token的dimnum_heads=8,qkv_bias=False,qk_scale=None,attn_drop_ratio=0.,proj_drop_ratio=0.):super(Attention, self).__init__()self.num_heads = num_headshead_dim = dim // num_headsself.scale = qk_scale or head_dim ** -0.5self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)self.attn_drop = nn.Dropout(attn_drop_ratio)self.proj = nn.Linear(dim, dim)self.proj_drop = nn.Dropout(proj_drop_ratio)def forward(self, x):# [batch_size, num_patches + 1, total_embed_dim]B, N, C = x.shape# qkv(): -> [batch_size, num_patches + 1, 3 * total_embed_dim]# reshape: -> [batch_size, num_patches + 1, 3, num_heads, embed_dim_per_head]# permute: -> [3, batch_size, num_heads, num_patches + 1, embed_dim_per_head]qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)# [batch_size, num_heads, num_patches + 1, embed_dim_per_head]q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)# transpose: -> [batch_size, num_heads, embed_dim_per_head, num_patches + 1]# @: multiply -> [batch_size, num_heads, num_patches + 1, num_patches + 1]attn = (q @ k.transpose(-2, -1)) * self.scaleattn = attn.softmax(dim=-1)attn = self.attn_drop(attn)# @: multiply -> [batch_size, num_heads, num_patches + 1, embed_dim_per_head]# transpose: -> [batch_size, num_patches + 1, num_heads, embed_dim_per_head]# reshape: -> [batch_size, num_patches + 1, total_embed_dim]x = (attn @ v).transpose(1, 2).reshape(B, N, C)x = self.proj(x)x = self.proj_drop(x)return xclass 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=nn.GELU, drop=0.):super().__init__()out_features = out_features or in_featureshidden_features = hidden_features or in_featuresself.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 xclass Block(nn.Module):def __init__(self,dim,num_heads,mlp_ratio=4.,qkv_bias=False,qk_scale=None,drop_ratio=0.,attn_drop_ratio=0.,drop_path_ratio=0.,act_layer=nn.GELU,norm_layer=nn.LayerNorm):super(Block, self).__init__()self.norm1 = norm_layer(dim)self.attn = Attention(dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale,attn_drop_ratio=attn_drop_ratio, proj_drop_ratio=drop_ratio)# NOTE: drop path for stochastic depth, we shall see if this is better than dropout hereself.drop_path = DropPath(drop_path_ratio) if drop_path_ratio > 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_ratio)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 xclass VisionTransformer(nn.Module):def __init__(self, img_size=224, patch_size=16, in_c=3, num_classes=1000,embed_dim=768, depth=12, num_heads=12, mlp_ratio=4.0, qkv_bias=True,qk_scale=None, representation_size=None, distilled=False, drop_ratio=0.,attn_drop_ratio=0., drop_path_ratio=0., embed_layer=PatchEmbed, norm_layer=None,act_layer=None):"""Args:img_size (int, tuple): input image sizepatch_size (int, tuple): patch sizein_c (int): number of input channelsnum_classes (int): number of classes for classification headembed_dim (int): embedding dimensiondepth (int): depth of transformer  Encoder Block的个数  L=12num_heads (int): number of attention headsmlp_ratio (int): ratio of mlp hidden dim to embedding dimqkv_bias (bool): enable bias for qkv if Trueqk_scale (float): override default qk scale of head_dim ** -0.5 if setrepresentation_size (Optional[int]): enable and set representation layer (pre-logits) to this value if set  pre-logits全连接层的节点个数distilled (bool): model includes a distillation token and head as in DeiT modelsdrop_ratio (float): dropout rateattn_drop_ratio (float): attention dropout ratedrop_path_ratio (float): stochastic depth rateembed_layer (nn.Module): patch embedding layernorm_layer: (nn.Module): normalization layer"""super(VisionTransformer, self).__init__()self.num_classes = num_classesself.num_features = self.embed_dim = embed_dim  # num_features for consistency with other modelsself.num_tokens = 2 if distilled else 1norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6)act_layer = act_layer or nn.GELUself.patch_embed = embed_layer(img_size=img_size, patch_size=patch_size, in_c=in_c, embed_dim=embed_dim)num_patches = self.patch_embed.num_patches# 1-batch 1 768self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))self.dist_token = nn.Parameter(torch.zeros(1, 1, embed_dim)) if distilled else None# num_patches 14×14 num_tokens 1self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + self.num_tokens, embed_dim))self.pos_drop = nn.Dropout(p=drop_ratio)# 构建等差序列(递增序列) 0~drop_path_ratio depth个元素dpr = [x.item() for x in torch.linspace(0, drop_path_ratio, depth)]  # stochastic depth decay rule# transformer encoder中encoder blockself.blocks = nn.Sequential(*[Block(dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,drop_ratio=drop_ratio, attn_drop_ratio=attn_drop_ratio, drop_path_ratio=dpr[i],norm_layer=norm_layer, act_layer=act_layer)for i in range(depth)])self.norm = norm_layer(embed_dim)# Representation layerif representation_size and not distilled:self.has_logits = Trueself.num_features = representation_sizeself.pre_logits = nn.Sequential(OrderedDict([("fc", nn.Linear(embed_dim, representation_size)),("act", nn.Tanh())]))else:self.has_logits = Falseself.pre_logits = nn.Identity()# Classifier head(s)self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity()self.head_dist = Noneif distilled:self.head_dist = nn.Linear(self.embed_dim, self.num_classes) if num_classes > 0 else nn.Identity()# Weight init 权重初始化nn.init.trunc_normal_(self.pos_embed, std=0.02)if self.dist_token is not None:nn.init.trunc_normal_(self.dist_token, std=0.02)nn.init.trunc_normal_(self.cls_token, std=0.02)self.apply(_init_vit_weights)def forward_features(self, x):# [B, C, H, W] -> [B, num_patches, embed_dim]# patch embeddingx = self.patch_embed(x)  # [B, 196, 768]# [1, 1, 768] -> [B, 1, 768]  复制batch_size份# Class token 1×768 与 196×768拼接 --> 197×768cls_token = self.cls_token.expand(x.shape[0], -1, -1)if self.dist_token is None:x = torch.cat((cls_token, x), dim=1)  # [B, 197, 768]  在196维度concat，拼接后为197else:x = torch.cat((cls_token, self.dist_token.expand(x.shape[0], -1, -1), x), dim=1)# Position Embedding 197×768x = self.pos_drop(x + self.pos_embed)# Transformer Encoder  Encoder Block L(×12)x = self.blocks(x)# Layer Norm 197×768x = self.norm(x)if self.dist_token is None:return self.pre_logits(x[:, 0])  # 取第一个维度batch所有数据，取第二个维度索引为0的数据else:return x[:, 0], x[:, 1]def forward(self, x):x = self.forward_features(x)if self.head_dist is not None:x, x_dist = self.head(x[0]), self.head_dist(x[1])if self.training and not torch.jit.is_scripting():# during inference, return the average of both classifier predictionsreturn x, x_distelse:return (x + x_dist) / 2else:# 对应最后Linear 全连接层x = self.head(x)return xdef _init_vit_weights(m):"""ViT weight initialization:param m: module"""if isinstance(m, nn.Linear):nn.init.trunc_normal_(m.weight, std=.01)if m.bias is not None:nn.init.zeros_(m.bias)elif isinstance(m, nn.Conv2d):nn.init.kaiming_normal_(m.weight, mode="fan_out")if m.bias is not None:nn.init.zeros_(m.bias)elif isinstance(m, nn.LayerNorm):nn.init.zeros_(m.bias)nn.init.ones_(m.weight)def vit_base_patch16_224(num_classes: int = 1000):"""ViT-Base model (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-1k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:链接: https://pan.baidu.com/s/1zqb08naP0RPqqfSXfkB2EA  密码: eu9f"""model = VisionTransformer(img_size=224,patch_size=16,embed_dim=768,depth=12,num_heads=12,representation_size=None,num_classes=num_classes)return model# num_classes 对应ImageNet-21K的类别个数为21843
def vit_base_patch16_224_in21k(num_classes: int = 21843, has_logits: bool = True):"""ViT-Base model (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch16_224_in21k-e5005f0a.pth"""model = VisionTransformer(img_size=224,patch_size=16,embed_dim=768,    # 对应Hidden sizedepth=12,         # Layersnum_heads=12,representation_size=768 if has_logits else None,num_classes=num_classes)return modeldef vit_base_patch32_224(num_classes: int = 1000):"""ViT-Base model (ViT-B/32) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-1k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:链接: https://pan.baidu.com/s/1hCv0U8pQomwAtHBYc4hmZg  密码: s5hl"""model = VisionTransformer(img_size=224,patch_size=32,embed_dim=768,depth=12,num_heads=12,representation_size=None,num_classes=num_classes)return modeldef vit_base_patch32_224_in21k(num_classes: int = 21843, has_logits: bool = True):"""ViT-Base model (ViT-B/32) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch32_224_in21k-8db57226.pth"""model = VisionTransformer(img_size=224,patch_size=32,embed_dim=768,depth=12,num_heads=12,representation_size=768 if has_logits else None,num_classes=num_classes)return modeldef vit_large_patch16_224(num_classes: int = 1000):"""ViT-Large model (ViT-L/16) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-1k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:链接: https://pan.baidu.com/s/1cxBgZJJ6qUWPSBNcE4TdRQ  密码: qqt8"""model = VisionTransformer(img_size=224,patch_size=16,embed_dim=1024,depth=24,num_heads=16,representation_size=None,num_classes=num_classes)return modeldef vit_large_patch16_224_in21k(num_classes: int = 21843, has_logits: bool = True):"""ViT-Large model (ViT-L/16) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch16_224_in21k-606da67d.pth"""model = VisionTransformer(img_size=224,patch_size=16,embed_dim=1024,depth=24,num_heads=16,representation_size=1024 if has_logits else None,num_classes=num_classes)return modeldef vit_large_patch32_224_in21k(num_classes: int = 21843, has_logits: bool = True):"""ViT-Large model (ViT-L/32) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer.weights ported from official Google JAX impl:https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch32_224_in21k-9046d2e7.pth"""model = VisionTransformer(img_size=224,patch_size=32,embed_dim=1024,depth=24,num_heads=16,representation_size=1024 if has_logits else None,num_classes=num_classes)return modeldef vit_huge_patch14_224_in21k(num_classes: int = 21843, has_logits: bool = True):"""ViT-Huge model (ViT-H/14) from original paper (https://arxiv.org/abs/2010.11929).ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer.NOTE: converted weights not currently available, too large for github release hosting."""model = VisionTransformer(img_size=224,patch_size=14,embed_dim=1280,depth=32,num_heads=16,representation_size=1280 if has_logits else None,num_classes=num_classes)return model

PyTorch深度学习（23）Transformer及网络结构ViT相关推荐

PyTorch深度学习（18）网络结构LeNet、AlexNet
CNN(Convolutional Neural Network) 目标分类 Classification 图像属于哪一类目标检索 Retrieval 相同种类归为一类目标检测 Dete ...
PyTorch深度学习（25）网络结构ConvNeXt
ConvNeXt 论文地址:https://arxiv.org/abs/2201.03545 一.改进点随着技术的不断发展,各种新的架构及优化策略促使Transformer拥有更好的效果相同策略训 ...
pytorch | 深度学习分割网络U-net的pytorch模型实现
原文:https://blog.csdn.net/u014722627/article/details/60883185 pytorch | 深度学习分割网络U-net的pytorch模型实现这个是 ...
【深度学习】Transformer在语义分割上的应用探索
[深度学习]Transformer在语义分割上的应用探索文章目录 1 Segmenter 2 Swin-Unet:Unet形状的纯Transformer的医学图像分割 3 复旦大学提出SETR:基于 ...
pytorch深度学习_深度学习和PyTorch的推荐系统实施
pytorch深度学习 The recommendation is a simple algorithm that works on the principle of data filtering. ...
pytorch深度学习入门笔记
Pytorch 深度学习入门笔记作者:梅如你学习来源: 公众号: 阿力阿哩哩.土堆碎念 B站视频:https://www.bilibili.com/video/BV1hE411t7RN? 中国大学 ...
PyTorch深度学习-跟着小土堆学习
目录学习视频链接一些问题 P4:Python/PyTorch学习中两大法宝函数-dir().help() P5:PyCharm及Jupyter使用及对比 P6:PyTorch加载数据初认识 P7: ...
【Pytorch深度学习实践】B站up刘二大人课程笔记——目录与索引（已完结）
从有代码的课程开始讨论 [Pytorch深度学习实践]B站up刘二大人之LinearModel -代码理解与实现(1/9) [Pytorch深度学习实践]B站up刘二大人之 Gradient Desc ...
《PyTorch 深度学习实践》第10讲卷积神经网络（基础篇）
文章目录 1 卷积层 1.1 torch.nn.Conv2d相关参数 1.2 填充:padding 1.3 步长:stride 2 最大池化层 3 手写数字识别该专栏内容为对该视频的学习记录:[&l ...
Pytorch深度学习实战教程：UNet语义分割网络
1 前言本文属于Pytorch深度学习语义分割系列教程. 该系列文章的内容有: Pytorch的基本使用语义分割算法讲解本文的开发环境如下: 开发环境:Windows 开发语言:Python3. ...

PyTorch深度学习（23）Transformer及网络结构ViT