YOLO算法改进Backbone系列之：EfficientViT

EfficientViT: Memory Effificient Vision Transformer with Cascaded Group Attention
摘要：视觉transformer由于其高模型能力而取得了巨大的成功。然而，它们卓越的性能伴随着沉重的计算成本，这使得它们不适合实时应用。在这篇论文中，本文提出了一个高速视觉transformer家族，名为EfficientViT。本文发现现有的transformer模型的速度通常受到内存低效操作的限制，特别是在MHSA中的张量重塑和单元函数。因此，本文设计了一种具有三明治布局的新构建块，即在高效FFN层之间使用单个内存绑定的MHSA，从而提高了内存效率，同时增强了信道通信。此外，本文发现注意图在头部之间具有很高的相似性，从而导致计算冗余。为了解决这个问题，本文提出了一个级联的群体注意模块，以不同的完整特征分割来馈送注意头，不仅节省了计算成本，而且提高了注意多样性。综合实验表明，高效vit优于现有的高效模型，在速度和精度之间取得了良好的平衡。例如，本文的EfficientViT-M5在准确率上比MobileNetV3-Large高出1.9%，而在Nvidia V100 GPU和Intel Xeon CPU上的吞吐量分别高出40.4%和45.2%。与最近的高效型号MobileViT-XXS相比，efficientvitt - m2的精度提高了1.8%，同时在GPU/CPU上运行速度提高了5.8 ×/3.7 ×，转换为ONNX格式时速度提高了7.4×

本文通过分析DeiT和Swin两个Transformer架构得出如下结论：

• 适当降低MHSA层利用率可以在提高模型性能的同时提高访存效率
• 在不同的头部使用不同的通道划分特征，而不是像MHSA那样对所有头部使用相同的全特征，可以有效地减少注意力计算冗余
• 典型的通道配置，即在每个阶段之后将通道数加倍或对所有块使用等效通道，可能在最后几个块中产生大量冗余
• 在维度相同的情况下，Q、K的冗余度比V大得多 a new building block with a sandwich
  layout（减少self-attention的次数）：之前是一个block self-attention->fc->self-attention->fc->self-attention->fc->…N次数；现在是一个blockfc->self-attention->fc；不仅能够提升内存效率而且能够增强通道间的计算
  cascaded group attention：让多头串联学习特征：第一个头学习完特征后，第二个头利用第一个头学习到的特征的基础上再去学习(原来的transformer是第二个头跟第一个头同时独立地去学习)，同理第三个头学习时也得利用上第二个头学习的结果再去学习

Efficientvit模型结构如下图所示：

a memory-efficient sandwich layout

Cascaded Group Attention：解决了原来模型中多头重复学习(学习到的特征很多都是相似的)的问题，这里每个头学到的特征都不同，而且越往下面的头学到的特征越丰富。

Q是主动查询的行为，特征比K更加丰富，所以额外做了个Token Interation
Q进行self-attention之前先通过多次分组卷积再一次学习
Parameter Reallocation
self-attention主要在进行QK，而且还需要对Q/K进行reshape，所以为了运算效率更快，Q与K的维度小一点
而V只在后面被QK得到的结果进行权重分配，没那么费劲，为了学习更多的特征，所以V维度更大一些

Efficientvit变体模型结构如下表所示：

在YOLOv5项目中添加EfficientViT模型作为Backbone使用的教程：
（1）将YOLOv5项目的models/yolo.py修改parse_model函数以及BaseModel的_forward_once函数

def parse_model(d, ch):  # model_dict, input_channels(3)# Parse a YOLOv5 model.yaml dictionaryLOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")anchors, nc, gd, gw, act = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'], d.get('activation')if act:Conv.default_act = eval(act)  # redefine default activation, i.e. Conv.default_act = nn.SiLU()LOGGER.info(f"{colorstr('activation:')} {act}")  # printna = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchorsno = na * (nc + 5)  # number of outputs = anchors * (classes + 5)# ---------------------------------------------------------------------------------------------------is_backbone = Falselayers, save, c2 = [], [], ch[-1]  # layers, savelist, ch outfor i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, argstry:t = mm = eval(m) if isinstance(m, str) else m  # eval stringsexcept:passfor j, a in enumerate(args):with contextlib.suppress(NameError):try:args[j] = eval(a) if isinstance(a, str) else a  # eval stringsexcept:args[j] = an = n_ = max(round(n * gd), 1) if n > 1 else n  # depth gainif m in {Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv,BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x}:c1, c2 = ch[f], args[0]if c2 != no:  # if not outputc2 = make_divisible(c2 * gw, 8)args = [c1, c2, *args[1:]]if m in {BottleneckCSP, C3, C3TR, C3Ghost, C3x}:args.insert(2, n)  # number of repeatsn = 1elif m is nn.BatchNorm2d:args = [ch[f]]elif m is Concat:c2 = sum(ch[x] for x in f)# TODO: channel, gw, gdelif m in {Detect, Segment}:args.append([ch[x] for x in f])if isinstance(args[1], int):  # number of anchorsargs[1] = [list(range(args[1] * 2))] * len(f)if m is Segment:args[3] = make_divisible(args[3] * gw, 8)elif m is Contract:c2 = ch[f] * args[0] ** 2elif m is Expand:c2 = ch[f] // args[0] ** 2# -------------------------------------------------------------------------------------elif m in {}:m = m(*args)c2 = m.channel# -------------------------------------------------------------------------------------else:c2 = ch[f]# -------------------------------------------------------------------------------------if isinstance(c2, list):is_backbone = Truem_ = mm_.backbone = Trueelse:m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # modulet = str(m)[8:-2].replace('__main__.', '')  # module type# -------------------------------------------------------------------------------------np = sum(x.numel() for x in m_.parameters())  # number params# -------------------------------------------------------------------------------------# m_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number paramsm_.i, m_.f, m_.type, m_.np = i + 4 if is_backbone else i, f, t, np  # attach index, 'from' index, type, number params# -------------------------------------------------------------------------------------LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # printsave.extend(x % (i + 4 if is_backbone else i) for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist# save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelistlayers.append(m_)if i == 0:ch = []# -------------------------------------------------------------------------------------if isinstance(c2, list):ch.extend(c2)for _ in range(5 - len(ch)):ch.insert(0, 0)else:ch.append(c2)# -------------------------------------------------------------------------------------return nn.Sequential(*layers), sorted(save)def _forward_once(self, x, profile=False, visualize=False):y, dt = [], []  # outputsfor m in self.model:if m.f != -1:  # if not from previous layerx = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layersif profile:self._profile_one_layer(m, x, dt)if hasattr(m, 'backbone'):x = m(x)for _ in range(5 - len(x)):x.insert(0, None)for i_idx, i in enumerate(x):if i_idx in self.save:y.append(i)else:y.append(None)x = x[-1]else:x = m(x)  # runy.append(x if m.i in self.save else None)  # save outputif visualize:feature_visualization(x, m.type, m.i, save_dir=visualize)return x

（2）在models/backbone（新建）文件下新建EfficientViT.py，添加如下的代码：

# --------------------------------------------------------
# EfficientViT Model Architecture for Downstream Tasks
# Copyright (c) 2022 Microsoft
# Written by: Xinyu Liu
# --------------------------------------------------------
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.checkpoint as checkpoint
import itertoolsfrom timm.models.layers import SqueezeExciteimport numpy as np
import itertools__all__ = ['EfficientViT_M0', 'EfficientViT_M1', 'EfficientViT_M2', 'EfficientViT_M3', 'EfficientViT_M4', 'EfficientViT_M5']class Conv2d_BN(torch.nn.Sequential):def __init__(self, a, b, ks=1, stride=1, pad=0, dilation=1,groups=1, bn_weight_init=1, resolution=-10000):super().__init__()self.add_module('c', torch.nn.Conv2d(a, b, ks, stride, pad, dilation, groups, bias=False))self.add_module('bn', torch.nn.BatchNorm2d(b))torch.nn.init.constant_(self.bn.weight, bn_weight_init)torch.nn.init.constant_(self.bn.bias, 0)@torch.no_grad()def fuse(self):c, bn = self._modules.values()w = bn.weight / (bn.running_var + bn.eps)**0.5w = c.weight * w[:, None, None, None]b = bn.bias - bn.running_mean * bn.weight / \(bn.running_var + bn.eps)**0.5m = torch.nn.Conv2d(w.size(1) * self.c.groups, w.size(0), w.shape[2:], stride=self.c.stride, padding=self.c.padding, dilation=self.c.dilation, groups=self.c.groups)m.weight.data.copy_(w)m.bias.data.copy_(b)return mdef replace_batchnorm(net):for child_name, child in net.named_children():if hasattr(child, 'fuse'):setattr(net, child_name, child.fuse())elif isinstance(child, torch.nn.BatchNorm2d):setattr(net, child_name, torch.nn.Identity())else:replace_batchnorm(child)class PatchMerging(torch.nn.Module):def __init__(self, dim, out_dim, input_resolution):super().__init__()hid_dim = int(dim * 4)self.conv1 = Conv2d_BN(dim, hid_dim, 1, 1, 0, resolution=input_resolution)self.act = torch.nn.ReLU()self.conv2 = Conv2d_BN(hid_dim, hid_dim, 3, 2, 1, groups=hid_dim, resolution=input_resolution)self.se = SqueezeExcite(hid_dim, .25)self.conv3 = Conv2d_BN(hid_dim, out_dim, 1, 1, 0, resolution=input_resolution // 2)def forward(self, x):x = self.conv3(self.se(self.act(self.conv2(self.act(self.conv1(x))))))return xclass Residual(torch.nn.Module):def __init__(self, m, drop=0.):super().__init__()self.m = mself.drop = dropdef forward(self, x):if self.training and self.drop > 0:return x + self.m(x) * torch.rand(x.size(0), 1, 1, 1,device=x.device).ge_(self.drop).div(1 - self.drop).detach()else:return x + self.m(x)class FFN(torch.nn.Module):def __init__(self, ed, h, resolution):super().__init__()self.pw1 = Conv2d_BN(ed, h, resolution=resolution)self.act = torch.nn.ReLU()self.pw2 = Conv2d_BN(h, ed, bn_weight_init=0, resolution=resolution)def forward(self, x):x = self.pw2(self.act(self.pw1(x)))return xclass CascadedGroupAttention(torch.nn.Module):r""" Cascaded Group Attention.Args:dim (int): Number of input channels.key_dim (int): The dimension for query and key.num_heads (int): Number of attention heads.attn_ratio (int): Multiplier for the query dim for value dimension.resolution (int): Input resolution, correspond to the window size.kernels (List[int]): The kernel size of the dw conv on query."""def __init__(self, dim, key_dim, num_heads=8,attn_ratio=4,resolution=14,kernels=[5, 5, 5, 5],):super().__init__()self.num_heads = num_headsself.scale = key_dim ** -0.5self.key_dim = key_dimself.d = int(attn_ratio * key_dim)self.attn_ratio = attn_ratioqkvs = []dws = []for i in range(num_heads):qkvs.append(Conv2d_BN(dim // (num_heads), self.key_dim * 2 + self.d, resolution=resolution))dws.append(Conv2d_BN(self.key_dim, self.key_dim, kernels[i], 1, kernels[i]//2, groups=self.key_dim, resolution=resolution))self.qkvs = torch.nn.ModuleList(qkvs)self.dws = torch.nn.ModuleList(dws)self.proj = torch.nn.Sequential(torch.nn.ReLU(), Conv2d_BN(self.d * num_heads, dim, bn_weight_init=0, resolution=resolution))points = list(itertools.product(range(resolution), range(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.abelse:self.ab = self.attention_biases[:, self.attention_bias_idxs]def forward(self, x):  # x (B,C,H,W)B, C, H, W = x.shapetrainingab = self.attention_biases[:, self.attention_bias_idxs]feats_in = x.chunk(len(self.qkvs), dim=1)feats_out = []feat = feats_in[0]for i, qkv in enumerate(self.qkvs):if i > 0: # add the previous output to the inputfeat = feat + feats_in[i]feat = qkv(feat)q, k, v = feat.view(B, -1, H, W).split([self.key_dim, self.key_dim, self.d], dim=1) # B, C/h, H, Wq = self.dws[i](q)q, k, v = q.flatten(2), k.flatten(2), v.flatten(2) # B, C/h, Nattn = ((q.transpose(-2, -1) @ k) * self.scale+(trainingab[i] if self.training else self.ab[i]))attn = attn.softmax(dim=-1) # BNNfeat = (v @ attn.transpose(-2, -1)).view(B, self.d, H, W) # BCHWfeats_out.append(feat)x = self.proj(torch.cat(feats_out, 1))return xclass LocalWindowAttention(torch.nn.Module):r""" Local Window Attention.Args:dim (int): Number of input channels.key_dim (int): The dimension for query and key.num_heads (int): Number of attention heads.attn_ratio (int): Multiplier for the query dim for value dimension.resolution (int): Input resolution.window_resolution (int): Local window resolution.kernels (List[int]): The kernel size of the dw conv on query."""def __init__(self, dim, key_dim, num_heads=8,attn_ratio=4,resolution=14,window_resolution=7,kernels=[5, 5, 5, 5],):super().__init__()self.dim = dimself.num_heads = num_headsself.resolution = resolutionassert window_resolution > 0, 'window_size must be greater than 0'self.window_resolution = window_resolutionself.attn = CascadedGroupAttention(dim, key_dim, num_heads,attn_ratio=attn_ratio, resolution=window_resolution,kernels=kernels,)def forward(self, x):B, C, H, W = x.shapeif H <= self.window_resolution and W <= self.window_resolution:x = self.attn(x)else:x = x.permute(0, 2, 3, 1)pad_b = (self.window_resolution - H %self.window_resolution) % self.window_resolutionpad_r = (self.window_resolution - W %self.window_resolution) % self.window_resolutionpadding = pad_b > 0 or pad_r > 0if padding:x = torch.nn.functional.pad(x, (0, 0, 0, pad_r, 0, pad_b))pH, pW = H + pad_b, W + pad_rnH = pH // self.window_resolutionnW = pW // self.window_resolution# window partition, BHWC -> B(nHh)(nWw)C -> BnHnWhwC -> (BnHnW)hwC -> (BnHnW)Chwx = x.view(B, nH, self.window_resolution, nW, self.window_resolution, C).transpose(2, 3).reshape(B * nH * nW, self.window_resolution, self.window_resolution, C).permute(0, 3, 1, 2)x = self.attn(x)# window reverse, (BnHnW)Chw -> (BnHnW)hwC -> BnHnWhwC -> B(nHh)(nWw)C -> BHWCx = x.permute(0, 2, 3, 1).view(B, nH, nW, self.window_resolution, self.window_resolution,C).transpose(2, 3).reshape(B, pH, pW, C)if padding:x = x[:, :H, :W].contiguous()x = x.permute(0, 3, 1, 2)return xclass EfficientViTBlock(torch.nn.Module):""" A basic EfficientViT building block.Args:type (str): Type for token mixer. Default: 's' for self-attention.ed (int): Number of input channels.kd (int): Dimension for query and key in the token mixer.nh (int): Number of attention heads.ar (int): Multiplier for the query dim for value dimension.resolution (int): Input resolution.window_resolution (int): Local window resolution.kernels (List[int]): The kernel size of the dw conv on query."""def __init__(self, type,ed, kd, nh=8,ar=4,resolution=14,window_resolution=7,kernels=[5, 5, 5, 5],):super().__init__()self.dw0 = Residual(Conv2d_BN(ed, ed, 3, 1, 1, groups=ed, bn_weight_init=0., resolution=resolution))self.ffn0 = Residual(FFN(ed, int(ed * 2), resolution))if type == 's':self.mixer = Residual(LocalWindowAttention(ed, kd, nh, attn_ratio=ar, \resolution=resolution, window_resolution=window_resolution, kernels=kernels))self.dw1 = Residual(Conv2d_BN(ed, ed, 3, 1, 1, groups=ed, bn_weight_init=0., resolution=resolution))self.ffn1 = Residual(FFN(ed, int(ed * 2), resolution))def forward(self, x):return self.ffn1(self.dw1(self.mixer(self.ffn0(self.dw0(x)))))class EfficientViT(torch.nn.Module):def __init__(self, img_size=400,patch_size=16,frozen_stages=0,in_chans=3,stages=['s', 's', 's'],embed_dim=[64, 128, 192],key_dim=[16, 16, 16],depth=[1, 2, 3],num_heads=[4, 4, 4],window_size=[7, 7, 7],kernels=[5, 5, 5, 5],down_ops=[['subsample', 2], ['subsample', 2], ['']],pretrained=None,distillation=False,):super().__init__()resolution = img_sizeself.patch_embed = torch.nn.Sequential(Conv2d_BN(in_chans, embed_dim[0] // 8, 3, 2, 1, resolution=resolution), torch.nn.ReLU(),Conv2d_BN(embed_dim[0] // 8, embed_dim[0] // 4, 3, 2, 1, resolution=resolution // 2), torch.nn.ReLU(),Conv2d_BN(embed_dim[0] // 4, embed_dim[0] // 2, 3, 2, 1, resolution=resolution // 4), torch.nn.ReLU(),Conv2d_BN(embed_dim[0] // 2, embed_dim[0], 3, 1, 1, resolution=resolution // 8))resolution = img_size // patch_sizeattn_ratio = [embed_dim[i] / (key_dim[i] * num_heads[i]) for i in range(len(embed_dim))]self.blocks1 = []self.blocks2 = []self.blocks3 = []for i, (stg, ed, kd, dpth, nh, ar, wd, do) in enumerate(zip(stages, embed_dim, key_dim, depth, num_heads, attn_ratio, window_size, down_ops)):for d in range(dpth):eval('self.blocks' + str(i+1)).append(EfficientViTBlock(stg, ed, kd, nh, ar, resolution, wd, kernels))if do[0] == 'subsample':#('Subsample' stride)blk = eval('self.blocks' + str(i+2))resolution_ = (resolution - 1) // do[1] + 1blk.append(torch.nn.Sequential(Residual(Conv2d_BN(embed_dim[i], embed_dim[i], 3, 1, 1, groups=embed_dim[i], resolution=resolution)),Residual(FFN(embed_dim[i], int(embed_dim[i] * 2), resolution)),))blk.append(PatchMerging(*embed_dim[i:i + 2], resolution))resolution = resolution_blk.append(torch.nn.Sequential(Residual(Conv2d_BN(embed_dim[i + 1], embed_dim[i + 1], 3, 1, 1, groups=embed_dim[i + 1], resolution=resolution)),Residual(FFN(embed_dim[i + 1], int(embed_dim[i + 1] * 2), resolution)),))self.blocks1 = torch.nn.Sequential(*self.blocks1)self.blocks2 = torch.nn.Sequential(*self.blocks2)self.blocks3 = torch.nn.Sequential(*self.blocks3)self.channel = [i.size(1) for i in self.forward(torch.randn(1, 3, 640, 640))]def forward(self, x):outs = []x = self.patch_embed(x)x = self.blocks1(x)outs.append(x)x = self.blocks2(x)outs.append(x)x = self.blocks3(x)outs.append(x)return outsEfficientViT_m0 = {'img_size': 224,'patch_size': 16,'embed_dim': [64, 128, 192],'depth': [1, 2, 3],'num_heads': [4, 4, 4],'window_size': [7, 7, 7],'kernels': [7, 5, 3, 3],}EfficientViT_m1 = {'img_size': 224,'patch_size': 16,'embed_dim': [128, 144, 192],'depth': [1, 2, 3],'num_heads': [2, 3, 3],'window_size': [7, 7, 7],'kernels': [7, 5, 3, 3],}EfficientViT_m2 = {'img_size': 224,'patch_size': 16,'embed_dim': [128, 192, 224],'depth': [1, 2, 3],'num_heads': [4, 3, 2],'window_size': [7, 7, 7],'kernels': [7, 5, 3, 3],}EfficientViT_m3 = {'img_size': 224,'patch_size': 16,'embed_dim': [128, 240, 320],'depth': [1, 2, 3],'num_heads': [4, 3, 4],'window_size': [7, 7, 7],'kernels': [5, 5, 5, 5],}EfficientViT_m4 = {'img_size': 224,'patch_size': 16,'embed_dim': [128, 256, 384],'depth': [1, 2, 3],'num_heads': [4, 4, 4],'window_size': [7, 7, 7],'kernels': [7, 5, 3, 3],}EfficientViT_m5 = {'img_size': 224,'patch_size': 16,'embed_dim': [192, 288, 384],'depth': [1, 3, 4],'num_heads': [3, 3, 4],'window_size': [7, 7, 7],'kernels': [7, 5, 3, 3],}def EfficientViT_M0(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m0):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef EfficientViT_M1(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m1):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef EfficientViT_M2(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m2):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef EfficientViT_M3(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m3):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef EfficientViT_M4(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m4):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef EfficientViT_M5(pretrained='', frozen_stages=0, distillation=False, fuse=False, pretrained_cfg=None, model_cfg=EfficientViT_m5):model = EfficientViT(frozen_stages=frozen_stages, distillation=distillation, pretrained=pretrained, **model_cfg)if pretrained:model.load_state_dict(update_weight(model.state_dict(), torch.load(pretrained)['model']))if fuse:replace_batchnorm(model)return modeldef update_weight(model_dict, weight_dict):idx, temp_dict = 0, {}for k, v in weight_dict.items():# k = k[9:]if k in model_dict.keys() and np.shape(model_dict[k]) == np.shape(v):temp_dict[k] = vidx += 1model_dict.update(temp_dict)print(f'loading weights... {idx}/{len(model_dict)} items')return model_dict

（3）在models/yolo.py导入EfficientViT模型并在parse_model函数中修改如下：

from models.backbone.EfficientViT import *
---------------------------------------------------
elif m in {EfficientViT_M0, EfficientViT_M1, EfficientViT_M2, EfficientViT_M3, EfficientViT_M4, EfficientViT_M5,}:
m = m(*args)
c2 = m.channel
---------------------------------------------------

（4）在model下面新建配置文件：yolov5-efficientvit.yaml

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license
# Parameters
nc: 80  # number of classes
depth_multiple: 0.33  # model depth multiple
width_multiple: 0.25  # layer channel multiple
anchors:- [10,13, 16,30, 33,23]  # P3/8- [30,61, 62,45, 59,119]  # P4/16- [116,90, 156,198, 373,326]  # P5/32# YOLOv5 v6.0 backbone
backbone:# [from, number, module, args][[-1, 1, EfficientViT_M0, []], # 4[-1, 1, SPPF, [1024, 5]],  # 5]# YOLOv5 v6.0 head
head:[[-1, 1, Conv, [512, 1, 1]], # 6[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 7[[-1, 3], 1, Concat, [1]],  # cat backbone P4 8[-1, 3, C3, [512, False]],  # 9[-1, 1, Conv, [256, 1, 1]], # 10[-1, 1, nn.Upsample, [None, 2, 'nearest']], # 11[[-1, 2], 1, Concat, [1]],  # cat backbone P3 12[-1, 3, C3, [256, False]],  # 13 (P3/8-small)[-1, 1, Conv, [256, 3, 2]], # 14[[-1, 10], 1, Concat, [1]],  # cat head P4 15[-1, 3, C3, [512, False]],  # 16 (P4/16-medium)[-1, 1, Conv, [512, 3, 2]], # 17[[-1, 5], 1, Concat, [1]],  # cat head P5 18[-1, 3, C3, [1024, False]],  # 19 (P5/32-large)[[13, 16, 19], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)]

（5）运行验证：在models/yolo.py文件指定–cfg参数为新建的yolov5-efficientvit.yaml

    from  n    params  module                                  arguments                     0                -1  1   2155680  EfficientViT_M0                         []                            1                -1  1    117440  models.common.SPPF                      [192, 256, 5]                 2                -1  1     33024  models.common.Conv                      [256, 128, 1, 1]              3                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          4           [-1, 3]  1         0  models.common.Concat                    [1]                           5                -1  1     90880  models.common.C3                        [256, 128, 1, False]          6                -1  1      8320  models.common.Conv                      [128, 64, 1, 1]               7                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          8           [-1, 2]  1         0  models.common.Concat                    [1]                           9                -1  1     22912  models.common.C3                        [128, 64, 1, False]           10                -1  1     36992  models.common.Conv                      [64, 64, 3, 2]                11          [-1, 10]  1         0  models.common.Concat                    [1]                           12                -1  1     74496  models.common.C3                        [128, 128, 1, False]          13                -1  1    147712  models.common.Conv                      [128, 128, 3, 2]              14           [-1, 5]  1         0  models.common.Concat                    [1]                           15                -1  1    329216  models.common.C3                        [384, 256, 1, False]          16      [13, 16, 19]  1    115005  Detect                                  [80, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [64, 128, 256]]
YOLOv5-efficientvit summary: 582 layers, 3131677 parameters, 3131677 gradients
Fusing layers... 
YOLOv5-efficientvit summary: 556 layers, 3129213 parameters, 3129213 gradients

目前整个项目计划更新至少有50+Vision Transformer Backbone，以及一些其他的改进策略，另外后续也会同步更新改进后的模型在MS COCO数据集上从零开始训练得到的模型权重和训练结果。想要了解项目的朋友私信博主或关注gzh：BestSongC 发送yolo改进即可获取项目信息。