当前位置: 首页 > article >正文

EfficientFormer实战:使用EfficientFormerV2实现图像分类任务(一)

摘要

EfficientFormerV2是一种通过重新思考ViT设计选择和引入细粒度联合搜索策略而开发出的新型移动视觉骨干网络。它结合了卷积和变换器的优势,通过一系列高效的设计改进和搜索方法,实现了在移动设备上既轻又快且保持高性能的目标。这一成果为在资源受限的硬件上有效部署视觉变换器模型提供了新的思路,其主要特点和优势可以概括如下:

特点

  1. 低延迟与高效参数

    • EfficientFormerV2通过一系列设计改进和搜索策略,实现了与MobileNet相当的低延迟和高效参数数量。在相同的模型大小和延迟约束下,其性能超越了传统轻量级CNN模型。
  2. 细粒度联合搜索策略

    • 该模型引入了细粒度的联合搜索策略,该策略同时优化模型的延迟和参数数量,从而找到在移动设备上高效运行的架构。
  3. 混合架构设计

    • EfficientFormerV2结合了卷积神经网络(CNNs)和变换器(ViT)的优势。它在早期阶段使用卷积来捕获局部信息,并在后期阶段使用多头自注意力(MHSA)来模拟全局依赖性。
  4. 令牌混合器与前馈网络改进

    • 使用深度卷积(DWCONV)替代传统的平均池化层作为令牌混合器,提高了模型性能而不引入额外延迟。同时,改进了前馈网络的设计。
  5. MHSA模块增强

    • 通过向值矩阵(V)注入局部信息,并在注意力头之间添加全连接层来增强MHSA模块的性能,进一步提升模型性能。
  6. 高效注意力机制应用

    • 提出了一种在高分辨率特征上有效应用MHSA的策略,通过下采样查询、键和值到固定分辨率,并插值回原始分辨率,从而在不显著增加延迟的情况下应用注意力机制。
  7. 双路径注意力下采样

    • 结合了静态局部下采样(如池化)和可学习的局部下采样(如深度卷积),形成双路径注意力下采样策略,提高了下采样过程的效率。
      在这里插入图片描述

优点

  1. 性能优越

    • 在ImageNet-1K等基准数据集上,EfficientFormerV2在相同或更低的模型大小和延迟下,表现出比MobileNet等轻量级CNN更高的准确率。
  2. 灵活性高

    • 通过细粒度的联合搜索策略,可以生成一系列具有不同模型大小和延迟的模型变体,以适应不同的移动设备和应用场景。
  3. 易于部署

    • 专为移动设备设计,具有较低的延迟和高效的参数数量,使得模型在实际应用中易于部署和推理。
  4. 兼容多种任务

    • EfficientFormerV2不仅在分类任务上表现出色,还可在目标检测、实例分割和语义分割等下游任务中作为骨干网络使用,并提升这些任务的性能。

本文使用EfficientFormerV2模型实现图像分类任务,模型选择efficientformerv2_s0,在植物幼苗分类任务ACC达到了96%+。

在这里插入图片描述
在这里插入图片描述

通过深入阅读本文,您将能够掌握以下关键技能与知识:

  1. 数据增强的多种策略:包括利用PyTorch的transforms库进行基本增强,以及进阶技巧如CutOut、MixUp、CutMix等,这些方法能显著提升模型泛化能力。

  2. GCViT模型的训练实现:了解如何从头开始构建并训练EfficientFormerV2(或其他深度学习模型),涵盖模型定义、数据加载、训练循环等关键环节。

  3. 混合精度训练:学习如何利用PyTorch自带的混合精度训练功能,加速训练过程同时减少内存消耗。

  4. 梯度裁剪技术:掌握梯度裁剪的应用,有效防止梯度爆炸问题,确保训练过程的稳定性。

  5. 分布式数据并行(DP)训练:了解如何在多GPU环境下使用PyTorch的分布式数据并行功能,加速大规模模型训练。

  6. 可视化训练过程:学习如何绘制训练过程中的loss和accuracy曲线,直观监控模型学习状况。

  7. 评估与生成报告:掌握在验证集上评估模型性能的方法,并生成详细的评估报告,包括ACC等指标。

  8. 测试脚本编写:学会编写测试脚本,对测试集进行预测,评估模型在实际应用中的表现。

  9. 学习率调整策略:理解并应用余弦退火策略动态调整学习率,优化训练效果。

  10. 自定义统计工具:使用AverageMeter类或其他工具统计和记录训练过程中的ACC、loss等关键指标,便于后续分析。

  11. 深入理解ACC1与ACC5:掌握图像分类任务中ACC1(Top-1准确率)和ACC5(Top-5准确率)的含义及其计算方法。

  12. 指数移动平均(EMA):学习如何在模型训练中应用EMA技术,进一步提升模型在测试集上的表现。

若您在以上任一领域基础尚浅,感到理解困难,推荐您参考我的专栏“经典主干网络精讲与实战”,该专栏从零开始,循序渐进地讲解上述所有知识点,助您轻松掌握深度学习中的这些核心技能。

安装包

安装timm

使用pip就行,命令:

pip install timm

mixup增强和EMA用到了timm

数据增强Cutout和Mixup

为了提高模型的泛化能力和性能,我在数据预处理阶段加入了Cutout和Mixup这两种数据增强技术。Cutout通过随机遮挡图像的一部分来强制模型学习更鲁棒的特征,而Mixup则通过混合两张图像及其标签来生成新的训练样本,从而增加数据的多样性。实现这两种增强需要安装torchtoolbox。安装命令:

pip install torchtoolbox

Cutout实现,在transforms中。

from torchtoolbox.transform import Cutout
# 数据预处理
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    Cutout(),
    transforms.ToTensor(),
    transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])

])

需要导入包:from timm.data.mixup import Mixup,

定义Mixup,和SoftTargetCrossEntropy

  mixup_fn = Mixup(
    mixup_alpha=0.8, cutmix_alpha=1.0, cutmix_minmax=None,
    prob=0.1, switch_prob=0.5, mode='batch',
    label_smoothing=0.1, num_classes=12)
 criterion_train = SoftTargetCrossEntropy()

Mixup 是一种在图像分类任务中常用的数据增强技术,它通过将两张图像以及其对应的标签进行线性组合来生成新的数据和标签。
参数详解:

mixup_alpha (float): mixup alpha 值,如果 > 0,则 mixup 处于活动状态。

cutmix_alpha (float):cutmix alpha 值,如果 > 0,cutmix 处于活动状态。

cutmix_minmax (List[float]):cutmix 最小/最大图像比率,cutmix 处于活动状态,如果不是 None,则使用这个 vs alpha。

如果设置了 cutmix_minmax 则cutmix_alpha 默认为1.0

prob (float): 每批次或元素应用 mixup 或 cutmix 的概率。

switch_prob (float): 当两者都处于活动状态时切换cutmix 和mixup 的概率 。

mode (str): 如何应用 mixup/cutmix 参数(每个’batch’,‘pair’(元素对),‘elem’(元素)。

correct_lam (bool): 当 cutmix bbox 被图像边框剪裁时应用。 lambda 校正

label_smoothing (float):将标签平滑应用于混合目标张量。

num_classes (int): 目标的类数。

EMA

EMA(Exponential Moving Average)在深度学习中是一种用于模型参数优化的技术,它通过计算参数的指数移动平均值来平滑模型的学习过程。这种方法有助于提高模型的稳定性和泛化能力,特别是在训练后期。以下是关于EMA的总结,表达进行了优化:

EMA概述

EMA是一种加权移动平均技术,其中每个新的平均值都是前一个平均值和当前值的加权和。在深度学习中,EMA被用于模型参数的更新,以减缓参数在训练过程中的快速波动,从而得到更加平滑和稳定的模型表现。

工作原理

在训练过程中,除了维护当前模型的参数外,还额外保存一份EMA参数。每个训练步骤或每隔一定步骤,根据当前模型参数和EMA参数,按照指数衰减的方式更新EMA参数。具体来说,EMA参数的更新公式通常如下:

EMA new = decay × EMA old + ( 1 − decay ) × model_parameters \text{EMA}_{\text{new}} = \text{decay} \times \text{EMA}_{\text{old}} + (1 - \text{decay}) \times \text{model\_parameters} EMAnew=decay×EMAold+(1decay)×model_parameters
其中,decay是一个介于0和1之间的超参数,控制着旧EMA值和新模型参数值之间的权重分配。较大的decay值意味着EMA更新时更多地依赖于旧值,即平滑效果更强。

应用优势

  1. 稳定性:EMA通过平滑参数更新过程,减少了模型在训练过程中的波动,使得模型更加稳定。
  2. 泛化能力:由于EMA参数是历史参数的平滑版本,它往往能捕捉到模型训练过程中的全局趋势,因此在测试或评估时,使用EMA参数往往能获得更好的泛化性能。
  3. 快速收敛:虽然EMA本身不直接加速训练过程,但通过稳定模型参数,它可能间接地帮助模型更快地收敛到更优的解。

使用场景

EMA在深度学习中的使用场景广泛,特别是在需要高度稳定性和良好泛化能力的任务中,如图像分类、目标检测等。在训练大型模型时,EMA尤其有用,因为它可以帮助减少过拟合的风险,并提高模型在未见数据上的表现。

具体实现如下:


import logging
from collections import OrderedDict
from copy import deepcopy
import torch
import torch.nn as nn

_logger = logging.getLogger(__name__)

class ModelEma:
    def __init__(self, model, decay=0.9999, device='', resume=''):
        # make a copy of the model for accumulating moving average of weights
        self.ema = deepcopy(model)
        self.ema.eval()
        self.decay = decay
        self.device = device  # perform ema on different device from model if set
        if device:
            self.ema.to(device=device)
        self.ema_has_module = hasattr(self.ema, 'module')
        if resume:
            self._load_checkpoint(resume)
        for p in self.ema.parameters():
            p.requires_grad_(False)

    def _load_checkpoint(self, checkpoint_path):
        checkpoint = torch.load(checkpoint_path, map_location='cpu')
        assert isinstance(checkpoint, dict)
        if 'state_dict_ema' in checkpoint:
            new_state_dict = OrderedDict()
            for k, v in checkpoint['state_dict_ema'].items():
                # ema model may have been wrapped by DataParallel, and need module prefix
                if self.ema_has_module:
                    name = 'module.' + k if not k.startswith('module') else k
                else:
                    name = k
                new_state_dict[name] = v
            self.ema.load_state_dict(new_state_dict)
            _logger.info("Loaded state_dict_ema")
        else:
            _logger.warning("Failed to find state_dict_ema, starting from loaded model weights")

    def update(self, model):
        # correct a mismatch in state dict keys
        needs_module = hasattr(model, 'module') and not self.ema_has_module
        with torch.no_grad():
            msd = model.state_dict()
            for k, ema_v in self.ema.state_dict().items():
                if needs_module:
                    k = 'module.' + k
                model_v = msd[k].detach()
                if self.device:
                    model_v = model_v.to(device=self.device)
                ema_v.copy_(ema_v * self.decay + (1. - self.decay) * model_v)

加入到模型中。

#初始化
if use_ema:
     model_ema = ModelEma(
            model_ft,
            decay=model_ema_decay,
            device='cpu',
            resume=resume)

# 训练过程中,更新完参数后,同步update shadow weights
def train():
    optimizer.step()
    if model_ema is not None:
        model_ema.update(model)


# 将model_ema传入验证函数中
val(model_ema.ema, DEVICE, test_loader)

针对没有预训练的模型,容易出现EMA不上分的情况,这点大家要注意啊!

项目结构

EfficientFormer_Demo
├─data1
│  ├─Black-grass
│  ├─Charlock
│  ├─Cleavers
│  ├─Common Chickweed
│  ├─Common wheat
│  ├─Fat Hen
│  ├─Loose Silky-bent
│  ├─Maize
│  ├─Scentless Mayweed
│  ├─Shepherds Purse
│  ├─Small-flowered Cranesbill
│  └─Sugar beet
├─models
│  └─efficientformer_v2.py
├─mean_std.py
├─makedata.py
├─train_timm.py
├─train.py
└─test.py

mean_std.py:计算mean和std的值。
makedata.py:生成数据集。
train.py:训练models文件下efficientformer_v2的模型
train_timm.py:训练timm库中的efficientformer_v2模型,timm库中的模型有预训练模型。
models:来源官方代码。

计算mean和std

在深度学习中,特别是在处理图像数据时,计算数据的均值(mean)和标准差(standard deviation, std)并进行归一化(Normalization)是加速模型收敛、提高模型性能的关键步骤之一。这里我将详细解释这两个概念,并讨论它们如何帮助模型学习。

均值(Mean)

均值是所有数值加和后除以数值的个数得到的平均值。在图像处理中,我们通常对每个颜色通道(如RGB图像的三个通道)分别计算均值。这意味着,如果我们的数据集包含多张图像,我们会计算所有图像在R通道上的像素值的均值,同样地,我们也会计算G通道和B通道的均值。

标准差(Standard Deviation, Std)

标准差是衡量数据分布离散程度的统计量。它反映了数据点与均值的偏离程度。在计算图像数据的标准差时,我们也是针对每个颜色通道分别进行的。标准差较大的颜色通道意味着该通道上的像素值变化较大,而标准差较小的通道则相对较为稳定。

归一化(Normalization)

归一化是将数据按比例缩放,使之落入一个小的特定区间,通常是[0, 1]或[-1, 1]。在图像处理中,我们通常会使用计算得到的均值和标准差来进行归一化,公式如下:

Normalized Value = Original Value − Mean Std \text{Normalized Value} = \frac{\text{Original Value} - \text{Mean}}{\text{Std}} Normalized Value=StdOriginal ValueMean

注意,在某些情况下,为了简化计算并确保数据非负,我们可能会选择将数据缩放到[0, 1]区间,这时使用的是最大最小值归一化,而不是基于均值和标准差的归一化。但在这里,我们主要讨论基于均值和标准差的归一化,因为它能保留数据的分布特性。

为什么需要归一化?

  1. 加速收敛:归一化后的数据具有相似的尺度,这有助于梯度下降算法更快地找到最优解,因为不同特征的梯度更新将在同一数量级上,从而避免了某些特征因尺度过大或过小而导致的训练缓慢或梯度消失/爆炸问题。

  2. 提高精度:归一化可以改善模型的泛化能力,因为它使得模型更容易学习到特征之间的相对关系,而不是被特征的绝对大小所影响。

  3. 稳定性:归一化后的数据更加稳定,减少了训练过程中的波动,有助于模型更加稳定地收敛。

如何计算和使用mean和std

  1. 计算全局mean和std:在整个数据集上计算mean和std。这通常是在训练开始前进行的,并使用这些值来归一化训练集、验证集和测试集。

  2. 使用库函数:许多深度学习框架(如PyTorch、TensorFlow等)提供了计算mean和std的便捷函数,并可以直接用于数据集的归一化。

  3. 动态调整:在某些情况下,特别是当数据集非常大或持续更新时,可能需要动态地计算mean和std。这通常涉及到在训练过程中使用移动平均(如EMA)来更新这些统计量。

计算并使用数据的mean和std进行归一化是深度学习中的一项基本且重要的预处理步骤,它对于加速模型收敛、提高模型性能和稳定性具有重要意义。新建mean_std.py,插入代码:

from torchvision.datasets import ImageFolder
import torch
from torchvision import transforms

def get_mean_and_std(train_data):
    train_loader = torch.utils.data.DataLoader(
        train_data, batch_size=1, shuffle=False, num_workers=0,
        pin_memory=True)
    mean = torch.zeros(3)
    std = torch.zeros(3)
    for X, _ in train_loader:
        for d in range(3):
            mean[d] += X[:, d, :, :].mean()
            std[d] += X[:, d, :, :].std()
    mean.div_(len(train_data))
    std.div_(len(train_data))
    return list(mean.numpy()), list(std.numpy())

if __name__ == '__main__':
    train_dataset = ImageFolder(root=r'data1', transform=transforms.ToTensor())
    print(get_mean_and_std(train_dataset))

数据集结构:

image-20220221153058619

运行结果:

([0.3281186, 0.28937867, 0.20702125], [0.09407319, 0.09732835, 0.106712654])

把这个结果记录下来,后面要用!

生成数据集

我们整理还的图像分类的数据集结构是这样的

data
├─Black-grass
├─Charlock
├─Cleavers
├─Common Chickweed
├─Common wheat
├─Fat Hen
├─Loose Silky-bent
├─Maize
├─Scentless Mayweed
├─Shepherds Purse
├─Small-flowered Cranesbill
└─Sugar beet

pytorch和keras默认加载方式是ImageNet数据集格式,格式是

├─data
│  ├─val
│  │   ├─Black-grass
│  │   ├─Charlock
│  │   ├─Cleavers
│  │   ├─Common Chickweed
│  │   ├─Common wheat
│  │   ├─Fat Hen
│  │   ├─Loose Silky-bent
│  │   ├─Maize
│  │   ├─Scentless Mayweed
│  │   ├─Shepherds Purse
│  │   ├─Small-flowered Cranesbill
│  │   └─Sugar beet
│  └─train
│      ├─Black-grass
│      ├─Charlock
│      ├─Cleavers
│      ├─Common Chickweed
│      ├─Common wheat
│      ├─Fat Hen
│      ├─Loose Silky-bent
│      ├─Maize
│      ├─Scentless Mayweed
│      ├─Shepherds Purse
│      ├─Small-flowered Cranesbill
│      └─Sugar beet

新增格式转化脚本makedata.py,插入代码:

import glob
import os
import shutil

image_list=glob.glob('data1/*/*.png')
print(image_list)
file_dir='data'
if os.path.exists(file_dir):
    print('true')
    #os.rmdir(file_dir)
    shutil.rmtree(file_dir)#删除再建立
    os.makedirs(file_dir)
else:
    os.makedirs(file_dir)

from sklearn.model_selection import train_test_split
trainval_files, val_files = train_test_split(image_list, test_size=0.3, random_state=42)
train_dir='train'
val_dir='val'
train_root=os.path.join(file_dir,train_dir)
val_root=os.path.join(file_dir,val_dir)
for file in trainval_files:
    file_class=file.replace("\\","/").split('/')[-2]
    file_name=file.replace("\\","/").split('/')[-1]
    file_class=os.path.join(train_root,file_class)
    if not os.path.isdir(file_class):
        os.makedirs(file_class)
    shutil.copy(file, file_class + '/' + file_name)

for file in val_files:
    file_class=file.replace("\\","/").split('/')[-2]
    file_name=file.replace("\\","/").split('/')[-1]
    file_class=os.path.join(val_root,file_class)
    if not os.path.isdir(file_class):
        os.makedirs(file_class)
    shutil.copy(file, file_class + '/' + file_name)

完成上面的内容就可以开启训练和测试了。

EfficientFormerV2代码

"""
EfficientFormer_v2
"""
import os
import copy
import torch
import torch.nn as nn
import torch.nn.functional as F
import math
from typing import Dict
import itertools

from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD
from timm.models.layers import DropPath, trunc_normal_
from timm.models.registry import register_model
from timm.layers.helpers import to_2tuple

EfficientFormer_width = {
    'L': [40, 80, 192, 384],  # 26m 83.3% 6attn
    'S2': [32, 64, 144, 288],  # 12m 81.6% 4attn dp0.02
    'S1': [32, 48, 120, 224],  # 6.1m 79.0
    'S0': [32, 48, 96, 176],  # 75.0 75.7
}

EfficientFormer_depth = {
    'L': [5, 5, 15, 10],  # 26m 83.3%
    'S2': [4, 4, 12, 8],  # 12m
    'S1': [3, 3, 9, 6],  # 79.0
    'S0': [2, 2, 6, 4],  # 75.7
}

# 26m
expansion_ratios_L = {
    '0': [4, 4, 4, 4, 4],
    '1': [4, 4, 4, 4, 4],
    '2': [4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4],
    '3': [4, 4, 4, 3, 3, 3, 3, 4, 4, 4],
}

# 12m
expansion_ratios_S2 = {
    '0': [4, 4, 4, 4],
    '1': [4, 4, 4, 4],
    '2': [4, 4, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4],
    '3': [4, 4, 3, 3, 3, 3, 4, 4],
}

# 6.1m
expansion_ratios_S1 = {
    '0': [4, 4, 4],
    '1': [4, 4, 4],
    '2': [4, 4, 3, 3, 3, 3, 4, 4, 4],
    '3': [4, 4, 3, 3, 4, 4],
}

# 3.5m
expansion_ratios_S0 = {
    '0': [4, 4],
    '1': [4, 4],
    '2': [4, 3, 3, 3, 4, 4],
    '3': [4, 3, 3, 4],
}


class Attention4D(torch.nn.Module):
    def __init__(self, dim=384, key_dim=32, num_heads=8,
                 attn_ratio=4,
                 resolution=7,
                 act_layer=nn.ReLU,
                 stride=None):
        super().__init__()
        self.num_heads = num_heads
        self.scale = key_dim ** -0.5
        self.key_dim = key_dim
        self.nh_kd = nh_kd = key_dim * num_heads

        if stride is not None:
            self.resolution = math.ceil(resolution / stride)
            self.stride_conv = nn.Sequential(nn.Conv2d(dim, dim, kernel_size=3, stride=stride, padding=1, groups=dim),
                                             nn.BatchNorm2d(dim), )
            self.upsample = nn.Upsample(scale_factor=stride, mode='bilinear')
        else:
            self.resolution = resolution
            self.stride_conv = None
            self.upsample = None

        self.N = self.resolution ** 2
        self.N2 = self.N
        self.d = int(attn_ratio * key_dim)
        self.dh = int(attn_ratio * key_dim) * num_heads
        self.attn_ratio = attn_ratio
        h = self.dh + nh_kd * 2
        self.q = nn.Sequential(nn.Conv2d(dim, self.num_heads * self.key_dim, 1),
                               nn.BatchNorm2d(self.num_heads * self.key_dim), )
        self.k = nn.Sequential(nn.Conv2d(dim, self.num_heads * self.key_dim, 1),
                               nn.BatchNorm2d(self.num_heads * self.key_dim), )
        self.v = nn.Sequential(nn.Conv2d(dim, self.num_heads * self.d, 1),
                               nn.BatchNorm2d(self.num_heads * self.d),
                               )
        self.v_local = nn.Sequential(nn.Conv2d(self.num_heads * self.d, self.num_heads * self.d,
                                               kernel_size=3, stride=1, padding=1, groups=self.num_heads * self.d),
                                     nn.BatchNorm2d(self.num_heads * self.d), )
        self.talking_head1 = nn.Conv2d(self.num_heads, self.num_heads, kernel_size=1, stride=1, padding=0)
        self.talking_head2 = nn.Conv2d(self.num_heads, self.num_heads, kernel_size=1, stride=1, padding=0)

        self.proj = nn.Sequential(act_layer(),
                                  nn.Conv2d(self.dh, dim, 1),
                                  nn.BatchNorm2d(dim), )

        points = list(itertools.product(range(self.resolution), range(self.resolution)))
        N = len(points)
        attention_offsets = {}
        idxs = []
        for p1 in points:
            for p2 in points:
                offset = (abs(p1[0] - p2[0]), abs(p1[1] - p2[1]))
                if offset not in attention_offsets:
                    attention_offsets[offset] = len(attention_offsets)
                idxs.append(attention_offsets[offset])
        self.attention_biases = torch.nn.Parameter(
            torch.zeros(num_heads, len(attention_offsets)))
        self.register_buffer('attention_bias_idxs',
                             torch.LongTensor(idxs).view(N, N))

    @torch.no_grad()
    def train(self, mode=True):
        super().train(mode)
        if mode and hasattr(self, 'ab'):
            del self.ab
        else:
            self.ab = self.attention_biases[:, self.attention_bias_idxs]

    def forward(self, x):  # x (B,N,C)
        B, C, H, W = x.shape
        if self.stride_conv is not None:
            x = self.stride_conv(x)

        q = self.q(x).flatten(2).reshape(B, self.num_heads, -1, self.N).permute(0, 1, 3, 2)
        k = self.k(x).flatten(2).reshape(B, self.num_heads, -1, self.N).permute(0, 1, 2, 3)
        v = self.v(x)
        v_local = self.v_local(v)
        v = v.flatten(2).reshape(B, self.num_heads, -1, self.N).permute(0, 1, 3, 2)

        attn = (
                (q @ k) * self.scale
                +
                (self.attention_biases[:, self.attention_bias_idxs]
                 if self.training else self.ab)
        )
        # attn = (q @ k) * self.scale
        attn = self.talking_head1(attn)
        attn = attn.softmax(dim=-1)
        attn = self.talking_head2(attn)

        x = (attn @ v)

        out = x.transpose(2, 3).reshape(B, self.dh, self.resolution, self.resolution) + v_local
        if self.upsample is not None:
            out = self.upsample(out)

        out = self.proj(out)
        return out


def stem(in_chs, out_chs, act_layer=nn.ReLU):
    return nn.Sequential(
        nn.Conv2d(in_chs, out_chs // 2, kernel_size=3, stride=2, padding=1),
        nn.BatchNorm2d(out_chs // 2),
        act_layer(),
        nn.Conv2d(out_chs // 2, out_chs, kernel_size=3, stride=2, padding=1),
        nn.BatchNorm2d(out_chs),
        act_layer(),
    )


class LGQuery(torch.nn.Module):
    def __init__(self, in_dim, out_dim, resolution1, resolution2):
        super().__init__()
        self.resolution1 = resolution1
        self.resolution2 = resolution2
        self.pool = nn.AvgPool2d(1, 2, 0)
        self.local = nn.Sequential(nn.Conv2d(in_dim, in_dim, kernel_size=3, stride=2, padding=1, groups=in_dim),
                                   )
        self.proj = nn.Sequential(nn.Conv2d(in_dim, out_dim, 1),
                                  nn.BatchNorm2d(out_dim), )

    def forward(self, x):
        local_q = self.local(x)
        pool_q = self.pool(x)
        q = local_q + pool_q
        q = self.proj(q)
        return q


class Attention4DDownsample(torch.nn.Module):
    def __init__(self, dim=384, key_dim=16, num_heads=8,
                 attn_ratio=4,
                 resolution=7,
                 out_dim=None,
                 act_layer=None,
                 ):
        super().__init__()

        self.num_heads = num_heads
        self.scale = key_dim ** -0.5
        self.key_dim = key_dim
        self.nh_kd = nh_kd = key_dim * num_heads

        self.resolution = resolution

        self.d = int(attn_ratio * key_dim)
        self.dh = int(attn_ratio * key_dim) * num_heads
        self.attn_ratio = attn_ratio
        h = self.dh + nh_kd * 2

        if out_dim is not None:
            self.out_dim = out_dim
        else:
            self.out_dim = dim
        self.resolution2 = math.ceil(self.resolution / 2)
        self.q = LGQuery(dim, self.num_heads * self.key_dim, self.resolution, self.resolution2)

        self.N = self.resolution ** 2
        self.N2 = self.resolution2 ** 2

        self.k = nn.Sequential(nn.Conv2d(dim, self.num_heads * self.key_dim, 1),
                               nn.BatchNorm2d(self.num_heads * self.key_dim), )
        self.v = nn.Sequential(nn.Conv2d(dim, self.num_heads * self.d, 1),
                               nn.BatchNorm2d(self.num_heads * self.d),
                               )
        self.v_local = nn.Sequential(nn.Conv2d(self.num_heads * self.d, self.num_heads * self.d,
                                               kernel_size=3, stride=2, padding=1, groups=self.num_heads * self.d),
                                     nn.BatchNorm2d(self.num_heads * self.d), )

        self.proj = nn.Sequential(
            act_layer(),
            nn.Conv2d(self.dh, self.out_dim, 1),
            nn.BatchNorm2d(self.out_dim), )

        points = list(itertools.product(range(self.resolution), range(self.resolution)))
        points_ = list(itertools.product(
            range(self.resolution2), range(self.resolution2)))
        N = len(points)
        N_ = len(points_)
        attention_offsets = {}
        idxs = []
        for p1 in points_:
            for p2 in points:
                size = 1
                offset = (
                    abs(p1[0] * math.ceil(self.resolution / self.resolution2) - p2[0] + (size - 1) / 2),
                    abs(p1[1] * math.ceil(self.resolution / self.resolution2) - p2[1] + (size - 1) / 2))
                if offset not in attention_offsets:
                    attention_offsets[offset] = len(attention_offsets)
                idxs.append(attention_offsets[offset])
        self.attention_biases = torch.nn.Parameter(
            torch.zeros(num_heads, len(attention_offsets)))
        self.register_buffer('attention_bias_idxs',
                             torch.LongTensor(idxs).view(N_, N))

    @torch.no_grad()
    def train(self, mode=True):
        super().train(mode)
        if mode and hasattr(self, 'ab'):
            del self.ab
        else:
            self.ab = self.attention_biases[:, self.attention_bias_idxs]

    def forward(self, x):  # x (B,N,C)
        B, C, H, W = x.shape

        q = self.q(x).flatten(2).reshape(B, self.num_heads, -1, self.N2).permute(0, 1, 3, 2)
        k = self.k(x).flatten(2).reshape(B, self.num_heads, -1, self.N).permute(0, 1, 2, 3)
        v = self.v(x)
        v_local = self.v_local(v)
        v = v.flatten(2).reshape(B, self.num_heads, -1, self.N).permute(0, 1, 3, 2)

        attn = (
                (q @ k) * self.scale
                +
                (self.attention_biases[:, self.attention_bias_idxs]
                 if self.training else self.ab)
        )

        # attn = (q @ k) * self.scale
        attn = attn.softmax(dim=-1)
        x = (attn @ v).transpose(2, 3)
        out = x.reshape(B, self.dh, self.resolution2, self.resolution2) + v_local

        out = self.proj(out)
        return out


class Embedding(nn.Module):
    def __init__(self, patch_size=3, stride=2, padding=1,
                 in_chans=3, embed_dim=768, norm_layer=nn.BatchNorm2d,
                 light=False, asub=False, resolution=None, act_layer=nn.ReLU, attn_block=Attention4DDownsample):
        super().__init__()
        self.light = light
        self.asub = asub

        if self.light:
            self.new_proj = nn.Sequential(
                nn.Conv2d(in_chans, in_chans, kernel_size=3, stride=2, padding=1, groups=in_chans),
                nn.BatchNorm2d(in_chans),
                nn.Hardswish(),
                nn.Conv2d(in_chans, embed_dim, kernel_size=1, stride=1, padding=0),
                nn.BatchNorm2d(embed_dim),
            )
            self.skip = nn.Sequential(
                nn.Conv2d(in_chans, embed_dim, kernel_size=1, stride=2, padding=0),
                nn.BatchNorm2d(embed_dim)
            )
        elif self.asub:
            self.attn = attn_block(dim=in_chans, out_dim=embed_dim,
                                   resolution=resolution, act_layer=act_layer)
            patch_size = to_2tuple(patch_size)
            stride = to_2tuple(stride)
            padding = to_2tuple(padding)
            self.conv = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size,
                                  stride=stride, padding=padding)
            self.bn = norm_layer(embed_dim) if norm_layer else nn.Identity()
        else:
            patch_size = to_2tuple(patch_size)
            stride = to_2tuple(stride)
            padding = to_2tuple(padding)
            self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size,
                                  stride=stride, padding=padding)
            self.norm = norm_layer(embed_dim) if norm_layer else nn.Identity()

    def forward(self, x):
        if self.light:
            out = self.new_proj(x) + self.skip(x)
        elif self.asub:
            out_conv = self.conv(x)
            out_conv = self.bn(out_conv)
            out = self.attn(x) + out_conv
        else:
            x = self.proj(x)
            out = self.norm(x)
        return out


class Mlp(nn.Module):
    """
    Implementation of MLP with 1*1 convolutions.
    Input: tensor with shape [B, C, H, W]
    """

    def __init__(self, in_features, hidden_features=None,
                 out_features=None, act_layer=nn.GELU, drop=0., mid_conv=False):
        super().__init__()
        out_features = out_features or in_features
        hidden_features = hidden_features or in_features
        self.mid_conv = mid_conv
        self.fc1 = nn.Conv2d(in_features, hidden_features, 1)
        self.act = act_layer()
        self.fc2 = nn.Conv2d(hidden_features, out_features, 1)
        self.drop = nn.Dropout(drop)
        self.apply(self._init_weights)

        if self.mid_conv:
            self.mid = nn.Conv2d(hidden_features, hidden_features, kernel_size=3, stride=1, padding=1,
                                 groups=hidden_features)
            self.mid_norm = nn.BatchNorm2d(hidden_features)

        self.norm1 = nn.BatchNorm2d(hidden_features)
        self.norm2 = nn.BatchNorm2d(out_features)

    def _init_weights(self, m):
        if isinstance(m, nn.Conv2d):
            trunc_normal_(m.weight, std=.02)
            if m.bias is not None:
                nn.init.constant_(m.bias, 0)

    def forward(self, x):
        x = self.fc1(x)
        x = self.norm1(x)
        x = self.act(x)

        if self.mid_conv:
            x_mid = self.mid(x)
            x_mid = self.mid_norm(x_mid)
            x = self.act(x_mid)
        x = self.drop(x)

        x = self.fc2(x)
        x = self.norm2(x)

        x = self.drop(x)
        return x


class AttnFFN(nn.Module):
    def __init__(self, dim, mlp_ratio=4.,
                 act_layer=nn.ReLU, norm_layer=nn.LayerNorm,
                 drop=0., drop_path=0.,
                 use_layer_scale=True, layer_scale_init_value=1e-5,
                 resolution=7, stride=None):

        super().__init__()

        self.token_mixer = Attention4D(dim, resolution=resolution, act_layer=act_layer, stride=stride)
        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim,
                       act_layer=act_layer, drop=drop, mid_conv=True)

        self.drop_path = DropPath(drop_path) if drop_path > 0. \
            else nn.Identity()
        self.use_layer_scale = use_layer_scale
        if use_layer_scale:
            self.layer_scale_1 = nn.Parameter(
                layer_scale_init_value * torch.ones(dim).unsqueeze(-1).unsqueeze(-1), requires_grad=True)
            self.layer_scale_2 = nn.Parameter(
                layer_scale_init_value * torch.ones(dim).unsqueeze(-1).unsqueeze(-1), requires_grad=True)

    def forward(self, x):
        if self.use_layer_scale:
            x = x + self.drop_path(self.layer_scale_1 * self.token_mixer(x))
            x = x + self.drop_path(self.layer_scale_2 * self.mlp(x))

        else:
            x = x + self.drop_path(self.token_mixer(x))
            x = x + self.drop_path(self.mlp(x))
        return x


class FFN(nn.Module):
    def __init__(self, dim, pool_size=3, mlp_ratio=4.,
                 act_layer=nn.GELU,
                 drop=0., drop_path=0.,
                 use_layer_scale=True, layer_scale_init_value=1e-5):
        super().__init__()

        mlp_hidden_dim = int(dim * mlp_ratio)
        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim,
                       act_layer=act_layer, drop=drop, mid_conv=True)

        self.drop_path = DropPath(drop_path) if drop_path > 0. \
            else nn.Identity()
        self.use_layer_scale = use_layer_scale
        if use_layer_scale:
            self.layer_scale_2 = nn.Parameter(
                layer_scale_init_value * torch.ones(dim).unsqueeze(-1).unsqueeze(-1), requires_grad=True)

    def forward(self, x):
        if self.use_layer_scale:
            x = x + self.drop_path(self.layer_scale_2 * self.mlp(x))
        else:
            x = x + self.drop_path(self.mlp(x))
        return x


def eformer_block(dim, index, layers,
                  pool_size=3, mlp_ratio=4.,
                  act_layer=nn.GELU, norm_layer=nn.LayerNorm,
                  drop_rate=.0, drop_path_rate=0.,
                  use_layer_scale=True, layer_scale_init_value=1e-5, vit_num=1, resolution=7, e_ratios=None):
    blocks = []
    for block_idx in range(layers[index]):
        block_dpr = drop_path_rate * (
                block_idx + sum(layers[:index])) / (sum(layers) - 1)
        mlp_ratio = e_ratios[str(index)][block_idx]
        if index >= 2 and block_idx > layers[index] - 1 - vit_num:
            if index == 2:
                stride = 2
            else:
                stride = None
            blocks.append(AttnFFN(
                dim, mlp_ratio=mlp_ratio,
                act_layer=act_layer, norm_layer=norm_layer,
                drop=drop_rate, drop_path=block_dpr,
                use_layer_scale=use_layer_scale,
                layer_scale_init_value=layer_scale_init_value,
                resolution=resolution,
                stride=stride,
            ))
        else:
            blocks.append(FFN(
                dim, pool_size=pool_size, mlp_ratio=mlp_ratio,
                act_layer=act_layer,
                drop=drop_rate, drop_path=block_dpr,
                use_layer_scale=use_layer_scale,
                layer_scale_init_value=layer_scale_init_value,
            ))
    blocks = nn.Sequential(*blocks)
    return blocks


class EfficientFormerV2(nn.Module):
    def __init__(self, layers, embed_dims=None,
                 mlp_ratios=4, downsamples=None,
                 pool_size=3,
                 norm_layer=nn.BatchNorm2d, act_layer=nn.GELU,
                 num_classes=1000,
                 down_patch_size=3, down_stride=2, down_pad=1,
                 drop_rate=0., drop_path_rate=0.,
                 use_layer_scale=True, layer_scale_init_value=1e-5,
                 fork_feat=False,
                 init_cfg=None,
                 pretrained=None,
                 vit_num=0,
                 distillation=True,
                 resolution=224,
                 e_ratios=expansion_ratios_L,
                 **kwargs):
        super().__init__()

        if not fork_feat:
            self.num_classes = num_classes
        self.fork_feat = fork_feat

        self.patch_embed = stem(3, embed_dims[0], act_layer=act_layer)

        network = []
        for i in range(len(layers)):
            stage = eformer_block(embed_dims[i], i, layers,
                                  pool_size=pool_size, mlp_ratio=mlp_ratios,
                                  act_layer=act_layer, norm_layer=norm_layer,
                                  drop_rate=drop_rate,
                                  drop_path_rate=drop_path_rate,
                                  use_layer_scale=use_layer_scale,
                                  layer_scale_init_value=layer_scale_init_value,
                                  resolution=math.ceil(resolution / (2 ** (i + 2))),
                                  vit_num=vit_num,
                                  e_ratios=e_ratios)
            network.append(stage)
            if i >= len(layers) - 1:
                break
            if downsamples[i] or embed_dims[i] != embed_dims[i + 1]:
                # downsampling between two stages
                if i >= 2:
                    asub = True
                else:
                    asub = False
                network.append(
                    Embedding(
                        patch_size=down_patch_size, stride=down_stride,
                        padding=down_pad,
                        in_chans=embed_dims[i], embed_dim=embed_dims[i + 1],
                        resolution=math.ceil(resolution / (2 ** (i + 2))),
                        asub=asub,
                        act_layer=act_layer, norm_layer=norm_layer,
                    )
                )

        self.network = nn.ModuleList(network)

        if self.fork_feat:
            # add a norm layer for each output
            self.out_indices = [0, 2, 4, 6]
            for i_emb, i_layer in enumerate(self.out_indices):
                if i_emb == 0 and os.environ.get('FORK_LAST3', None):
                    layer = nn.Identity()
                else:
                    layer = norm_layer(embed_dims[i_emb])
                layer_name = f'norm{i_layer}'
                self.add_module(layer_name, layer)
        else:
            # Classifier head
            self.norm = norm_layer(embed_dims[-1])
            self.head = nn.Linear(
                embed_dims[-1], num_classes) if num_classes > 0 \
                else nn.Identity()
            self.dist = distillation
            if self.dist:
                self.dist_head = nn.Linear(
                    embed_dims[-1], num_classes) if num_classes > 0 \
                    else nn.Identity()

        self.apply(self.cls_init_weights)

        self.init_cfg = copy.deepcopy(init_cfg)
        # load pre-trained model
        if self.fork_feat and (
                self.init_cfg is not None or pretrained is not None):
            self.init_weights()

    # init for classification
    def cls_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)

    # init for mmdetection or mmsegmentation by loading
    # imagenet pre-trained weights
    def init_weights(self, pretrained=None):
        logger = get_root_logger()
        if self.init_cfg is None and pretrained is None:
            logger.warn(f'No pre-trained weights for '
                        f'{self.__class__.__name__}, '
                        f'training start from scratch')
            pass
        else:
            assert 'checkpoint' in self.init_cfg, f'Only support ' \
                                                  f'specify `Pretrained` in ' \
                                                  f'`init_cfg` in ' \
                                                  f'{self.__class__.__name__} '
            if self.init_cfg is not None:
                ckpt_path = self.init_cfg['checkpoint']
            elif pretrained is not None:
                ckpt_path = pretrained

            ckpt = _load_checkpoint(
                ckpt_path, logger=logger, map_location='cpu')
            if 'state_dict' in ckpt:
                _state_dict = ckpt['state_dict']
            elif 'model' in ckpt:
                _state_dict = ckpt['model']
            else:
                _state_dict = ckpt

            state_dict = _state_dict
            missing_keys, unexpected_keys = \
                self.load_state_dict(state_dict, False)

    def forward_tokens(self, x):
        outs = []
        for idx, block in enumerate(self.network):
            x = block(x)
            if self.fork_feat and idx in self.out_indices:
                norm_layer = getattr(self, f'norm{idx}')
                x_out = norm_layer(x)
                outs.append(x_out)
        if self.fork_feat:
            return outs
        return x

    def forward(self, x):
        x = self.patch_embed(x)
        x = self.forward_tokens(x)
        if self.fork_feat:
            # otuput features of four stages for dense prediction
            return x
        # print(x.size())
        x = self.norm(x)
        if self.dist:
            cls_out = self.head(x.flatten(2).mean(-1)), self.dist_head(x.flatten(2).mean(-1))
            if not self.training:
                cls_out = (cls_out[0] + cls_out[1]) / 2
        else:
            cls_out = self.head(x.flatten(2).mean(-1))
        # for image classification
        return cls_out


def _cfg(url='', **kwargs):
    return {
        'url': url,
        'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': None,
        'crop_pct': .95, 'interpolation': 'bicubic',
        'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD,
        'classifier': 'head',
        **kwargs
    }


@register_model
def efficientformerv2_s0(pretrained=False, **kwargs):
    model = EfficientFormerV2(
        layers=EfficientFormer_depth['S0'],
        embed_dims=EfficientFormer_width['S0'],
        downsamples=[True, True, True, True, True],
        vit_num=2,
        drop_path_rate=0.0,
        e_ratios=expansion_ratios_S0,
        **kwargs)
    model.default_cfg = _cfg(crop_pct=0.9)
    return model


@register_model
def efficientformerv2_s1(pretrained=False, **kwargs):
    model = EfficientFormerV2(
        layers=EfficientFormer_depth['S1'],
        embed_dims=EfficientFormer_width['S1'],
        downsamples=[True, True, True, True],
        vit_num=2,
        drop_path_rate=0.0,
        e_ratios=expansion_ratios_S1,
        **kwargs)
    model.default_cfg = _cfg(crop_pct=0.9)
    return model


@register_model
def efficientformerv2_s2(pretrained=False, **kwargs):
    model = EfficientFormerV2(
        layers=EfficientFormer_depth['S2'],
        embed_dims=EfficientFormer_width['S2'],
        downsamples=[True, True, True, True],
        vit_num=4,
        drop_path_rate=0.02,
        e_ratios=expansion_ratios_S2,
        **kwargs)
    model.default_cfg = _cfg(crop_pct=0.9)
    return model


@register_model
def efficientformerv2_l(pretrained=False, **kwargs):
    model = EfficientFormerV2(
        layers=EfficientFormer_depth['L'],
        embed_dims=EfficientFormer_width['L'],
        downsamples=[True, True, True, True],
        vit_num=6,
        drop_path_rate=0.1,
        e_ratios=expansion_ratios_L,
        **kwargs)
    model.default_cfg = _cfg(crop_pct=0.9)
    return model


http://www.kler.cn/a/318183.html

相关文章:

  • TCP/IP 模型中,网络层对 IP 地址的分配与路由选择
  • Android U 多任务启动分屏——system_server流程(更新中)
  • CSS系列(27)- 图形与滤镜详解
  • WebDavClient 安装和配置指南
  • 设置中 wifi密码框被输入键盘遮挡的处理
  • 蓝桥杯物联网开发板硬件组成
  • WPF 的TreeView的TreeViewItem下动态生成TreeViewItem
  • 合宙LuatOS应用,与时间相关那些事
  • k8s中pod的创建过程和阶段状态
  • Allegro视频去除走线的小方块
  • Milvus - 四种一致性级别与应用场景解析
  • 可靠传输是什么?是基于UDP实现的吗
  • JUC并发编程_四大函数式接口和 Stream 流式计算
  • 适用于 Windows 的 7 大数据恢复工具,可靠的数据恢复工具可有效地恢复丢失的文件
  • 后端开发工程师转行大模型领域:全面学习路线指南,非常详细收藏我这一篇就够了
  • 【大语言模型_1】VLLM部署Qwen模型
  • 【速成Redis】04 Redis 概念扫盲:事务、持久化、主从复制、哨兵模式
  • 2-102基于matlab的蒙特卡洛仿真
  • C语言——文件操作
  • [数据结构]动态顺序表的实现与应用
  • 第二证券:“产业+科技” 中国并购重组市场持续升温
  • 【微服务即时通讯系统】——etcd一致性键值存储系统,etcd的介绍,etcd的安装,etcd使用和功能测试
  • Scikit-learn 识别手写数字
  • Qt:NULL与nullptr的区别(手写nullptr)
  • 数据处理与统计分析篇-day10-Matplotlib数据可视化
  • Leetcode 每日一题:Diameter of Binary Tree