YOLOv9大幅度按比例减小模型计算量！加快训练！

一、代码及论文链接：

代码链接：GitHub - WongKinYiu/yolov9: Implementation of paper - YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information
论文链接：https://github.com/WongKinYiu/yolov9/tree/main
二、说明

本文方法并不能直接替代YOLOv9原作者尚未开源的两个小模型，但可以按比例减小模型尺寸。类似YOLOv5、v8等，可以方便测试YOLOv9在数据集上的性能！方法来源于网络。
三、使用步骤

参照之前的YOLOv9代码，我们运行yolov9-c.yaml的参数量是239 GLOPs。

我们将以下代码替换掉YOLOv9工程下models包下yolo.py脚本中的代码。

import argparse
import os
import platform
import sys
from copy import deepcopy
from pathlib import PathFILE = Path(__file__).resolve()
ROOT = FILE.parents[1]  # YOLO root directory
if str(ROOT) not in sys.path:sys.path.append(str(ROOT))  # add ROOT to PATH
if platform.system() != 'Windows':ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relativefrom models.common import *
from models.experimental import *
from utils.general import LOGGER, check_version, check_yaml, make_divisible, print_args
from utils.plots import feature_visualization
from utils.torch_utils import (fuse_conv_and_bn, initialize_weights, model_info, profile, scale_img, select_device,time_sync)
from utils.tal.anchor_generator import make_anchors, dist2bboxtry:import thop  # for FLOPs computation
except ImportError:thop = Noneclass Detect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch)  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = max((ch[0] // 4, self.reg_max * 4, 16)), max((ch[0], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch)self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch)self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity()def forward(self, x):shape = x[0].shape  # BCHWfor i in range(self.nl):x[i] = torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1)if self.training:return xelif self.dynamic or self.shape != shape:self.anchors, self.strides = (x.transpose(0, 1) for x in make_anchors(x, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([xi.view(shape[0], self.no, -1) for xi in x], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesy = torch.cat((dbox, cls.sigmoid()), 1)return y if self.export else (y, x)def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class DDetect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch)  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = make_divisible(max((ch[0] // 4, self.reg_max * 4, 16)), 4), max((ch[0], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3, g=4), nn.Conv2d(c2, 4 * self.reg_max, 1, groups=4)) for x in ch)self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch)self.dfl = DFL(self.reg_max) if self.reg_max > 1 else nn.Identity()def forward(self, x):shape = x[0].shape  # BCHWfor i in range(self.nl):x[i] = torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1)if self.training:return xelif self.dynamic or self.shape != shape:self.anchors, self.strides = (x.transpose(0, 1) for x in make_anchors(x, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([xi.view(shape[0], self.no, -1) for xi in x], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesy = torch.cat((dbox, cls.sigmoid()), 1)return y if self.export else (y, x)def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class DualDetect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch) // 2  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = max((ch[0] // 4, self.reg_max * 4, 16)), max((ch[0], min((self.nc * 2, 128))))  # channelsc4, c5 = max((ch[self.nl] // 4, self.reg_max * 4, 16)), max((ch[self.nl], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch[:self.nl])self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch[:self.nl])self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, 4 * self.reg_max, 1)) for x in ch[self.nl:])self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nc, 1)) for x in ch[self.nl:])self.dfl = DFL(self.reg_max)self.dfl2 = DFL(self.reg_max)def forward(self, x):shape = x[0].shape  # BCHWd1 = []d2 = []for i in range(self.nl):d1.append(torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1))d2.append(torch.cat((self.cv4[i](x[self.nl+i]), self.cv5[i](x[self.nl+i])), 1))if self.training:return [d1, d2]elif self.dynamic or self.shape != shape:self.anchors, self.strides = (d1.transpose(0, 1) for d1 in make_anchors(d1, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([di.view(shape[0], self.no, -1) for di in d1], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox2, cls2 = torch.cat([di.view(shape[0], self.no, -1) for di in d2], 2).split((self.reg_max * 4, self.nc), 1)dbox2 = dist2bbox(self.dfl2(box2), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesy = [torch.cat((dbox, cls.sigmoid()), 1), torch.cat((dbox2, cls2.sigmoid()), 1)]return y if self.export else (y, [d1, d2])def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv4, m.cv5, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class DualDDetect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch) // 2  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = make_divisible(max((ch[0] // 4, self.reg_max * 4, 16)), 4), max((ch[0], min((self.nc * 2, 128))))  # channelsc4, c5 = make_divisible(max((ch[self.nl] // 4, self.reg_max * 4, 16)), 4), max((ch[self.nl], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3, g=4), nn.Conv2d(c2, 4 * self.reg_max, 1, groups=4)) for x in ch[:self.nl])self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch[:self.nl])self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3, g=4), nn.Conv2d(c4, 4 * self.reg_max, 1, groups=4)) for x in ch[self.nl:])self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nc, 1)) for x in ch[self.nl:])self.dfl = DFL(self.reg_max)self.dfl2 = DFL(self.reg_max)def forward(self, x):shape = x[0].shape  # BCHWd1 = []d2 = []for i in range(self.nl):d1.append(torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1))d2.append(torch.cat((self.cv4[i](x[self.nl+i]), self.cv5[i](x[self.nl+i])), 1))if self.training:return [d1, d2]elif self.dynamic or self.shape != shape:self.anchors, self.strides = (d1.transpose(0, 1) for d1 in make_anchors(d1, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([di.view(shape[0], self.no, -1) for di in d1], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox2, cls2 = torch.cat([di.view(shape[0], self.no, -1) for di in d2], 2).split((self.reg_max * 4, self.nc), 1)dbox2 = dist2bbox(self.dfl2(box2), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesy = [torch.cat((dbox, cls.sigmoid()), 1), torch.cat((dbox2, cls2.sigmoid()), 1)]return y if self.export else (y, [d1, d2])#y = torch.cat((dbox2, cls2.sigmoid()), 1)#return y if self.export else (y, d2)#y1 = torch.cat((dbox, cls.sigmoid()), 1)#y2 = torch.cat((dbox2, cls2.sigmoid()), 1)#return [y1, y2] if self.export else [(y1, d1), (y2, d2)]#return [y1, y2] if self.export else [(y1, y2), (d1, d2)]def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv4, m.cv5, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class TripleDetect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch) // 3  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = max((ch[0] // 4, self.reg_max * 4, 16)), max((ch[0], min((self.nc * 2, 128))))  # channelsc4, c5 = max((ch[self.nl] // 4, self.reg_max * 4, 16)), max((ch[self.nl], min((self.nc * 2, 128))))  # channelsc6, c7 = max((ch[self.nl * 2] // 4, self.reg_max * 4, 16)), max((ch[self.nl * 2], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3), nn.Conv2d(c2, 4 * self.reg_max, 1)) for x in ch[:self.nl])self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch[:self.nl])self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, 4 * self.reg_max, 1)) for x in ch[self.nl:self.nl*2])self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nc, 1)) for x in ch[self.nl:self.nl*2])self.cv6 = nn.ModuleList(nn.Sequential(Conv(x, c6, 3), Conv(c6, c6, 3), nn.Conv2d(c6, 4 * self.reg_max, 1)) for x in ch[self.nl*2:self.nl*3])self.cv7 = nn.ModuleList(nn.Sequential(Conv(x, c7, 3), Conv(c7, c7, 3), nn.Conv2d(c7, self.nc, 1)) for x in ch[self.nl*2:self.nl*3])self.dfl = DFL(self.reg_max)self.dfl2 = DFL(self.reg_max)self.dfl3 = DFL(self.reg_max)def forward(self, x):shape = x[0].shape  # BCHWd1 = []d2 = []d3 = []for i in range(self.nl):d1.append(torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1))d2.append(torch.cat((self.cv4[i](x[self.nl+i]), self.cv5[i](x[self.nl+i])), 1))d3.append(torch.cat((self.cv6[i](x[self.nl*2+i]), self.cv7[i](x[self.nl*2+i])), 1))if self.training:return [d1, d2, d3]elif self.dynamic or self.shape != shape:self.anchors, self.strides = (d1.transpose(0, 1) for d1 in make_anchors(d1, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([di.view(shape[0], self.no, -1) for di in d1], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox2, cls2 = torch.cat([di.view(shape[0], self.no, -1) for di in d2], 2).split((self.reg_max * 4, self.nc), 1)dbox2 = dist2bbox(self.dfl2(box2), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox3, cls3 = torch.cat([di.view(shape[0], self.no, -1) for di in d3], 2).split((self.reg_max * 4, self.nc), 1)dbox3 = dist2bbox(self.dfl3(box3), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesy = [torch.cat((dbox, cls.sigmoid()), 1), torch.cat((dbox2, cls2.sigmoid()), 1), torch.cat((dbox3, cls3.sigmoid()), 1)]return y if self.export else (y, [d1, d2, d3])def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv4, m.cv5, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv6, m.cv7, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class TripleDDetect(nn.Module):# YOLO Detect head for detection modelsdynamic = False  # force grid reconstructionexport = False  # export modeshape = Noneanchors = torch.empty(0)  # initstrides = torch.empty(0)  # initdef __init__(self, nc=80, ch=(), inplace=True):  # detection layersuper().__init__()self.nc = nc  # number of classesself.nl = len(ch) // 3  # number of detection layersself.reg_max = 16self.no = nc + self.reg_max * 4  # number of outputs per anchorself.inplace = inplace  # use inplace ops (e.g. slice assignment)self.stride = torch.zeros(self.nl)  # strides computed during buildc2, c3 = make_divisible(max((ch[0] // 4, self.reg_max * 4, 16)), 4), \max((ch[0], min((self.nc * 2, 128))))  # channelsc4, c5 = make_divisible(max((ch[self.nl] // 4, self.reg_max * 4, 16)), 4), \max((ch[self.nl], min((self.nc * 2, 128))))  # channelsc6, c7 = make_divisible(max((ch[self.nl * 2] // 4, self.reg_max * 4, 16)), 4), \max((ch[self.nl * 2], min((self.nc * 2, 128))))  # channelsself.cv2 = nn.ModuleList(nn.Sequential(Conv(x, c2, 3), Conv(c2, c2, 3, g=4), nn.Conv2d(c2, 4 * self.reg_max, 1, groups=4)) for x in ch[:self.nl])self.cv3 = nn.ModuleList(nn.Sequential(Conv(x, c3, 3), Conv(c3, c3, 3), nn.Conv2d(c3, self.nc, 1)) for x in ch[:self.nl])self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3, g=4), nn.Conv2d(c4, 4 * self.reg_max, 1, groups=4)) for x in ch[self.nl:self.nl*2])self.cv5 = nn.ModuleList(nn.Sequential(Conv(x, c5, 3), Conv(c5, c5, 3), nn.Conv2d(c5, self.nc, 1)) for x in ch[self.nl:self.nl*2])self.cv6 = nn.ModuleList(nn.Sequential(Conv(x, c6, 3), Conv(c6, c6, 3, g=4), nn.Conv2d(c6, 4 * self.reg_max, 1, groups=4)) for x in ch[self.nl*2:self.nl*3])self.cv7 = nn.ModuleList(nn.Sequential(Conv(x, c7, 3), Conv(c7, c7, 3), nn.Conv2d(c7, self.nc, 1)) for x in ch[self.nl*2:self.nl*3])self.dfl = DFL(self.reg_max)self.dfl2 = DFL(self.reg_max)self.dfl3 = DFL(self.reg_max)def forward(self, x):shape = x[0].shape  # BCHWd1 = []d2 = []d3 = []for i in range(self.nl):d1.append(torch.cat((self.cv2[i](x[i]), self.cv3[i](x[i])), 1))d2.append(torch.cat((self.cv4[i](x[self.nl+i]), self.cv5[i](x[self.nl+i])), 1))d3.append(torch.cat((self.cv6[i](x[self.nl*2+i]), self.cv7[i](x[self.nl*2+i])), 1))if self.training:return [d1, d2, d3]elif self.dynamic or self.shape != shape:self.anchors, self.strides = (d1.transpose(0, 1) for d1 in make_anchors(d1, self.stride, 0.5))self.shape = shapebox, cls = torch.cat([di.view(shape[0], self.no, -1) for di in d1], 2).split((self.reg_max * 4, self.nc), 1)dbox = dist2bbox(self.dfl(box), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox2, cls2 = torch.cat([di.view(shape[0], self.no, -1) for di in d2], 2).split((self.reg_max * 4, self.nc), 1)dbox2 = dist2bbox(self.dfl2(box2), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.stridesbox3, cls3 = torch.cat([di.view(shape[0], self.no, -1) for di in d3], 2).split((self.reg_max * 4, self.nc), 1)dbox3 = dist2bbox(self.dfl3(box3), self.anchors.unsqueeze(0), xywh=True, dim=1) * self.strides#y = [torch.cat((dbox, cls.sigmoid()), 1), torch.cat((dbox2, cls2.sigmoid()), 1), torch.cat((dbox3, cls3.sigmoid()), 1)]#return y if self.export else (y, [d1, d2, d3])y = torch.cat((dbox3, cls3.sigmoid()), 1)return y if self.export else (y, d3)def bias_init(self):# Initialize Detect() biases, WARNING: requires stride availabilitym = self  # self.model[-1]  # Detect() module# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1# ncf = math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # nominal class frequencyfor a, b, s in zip(m.cv2, m.cv3, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv4, m.cv5, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)for a, b, s in zip(m.cv6, m.cv7, m.stride):  # froma[-1].bias.data[:] = 1.0  # boxb[-1].bias.data[:m.nc] = math.log(5 / m.nc / (640 / s) ** 2)  # cls (5 objects and 80 classes per 640 image)class Segment(Detect):# YOLO Segment head for segmentation modelsdef __init__(self, nc=80, nm=32, npr=256, ch=(), inplace=True):super().__init__(nc, ch, inplace)self.nm = nm  # number of masksself.npr = npr  # number of protosself.proto = Proto(ch[0], self.npr, self.nm)  # protosself.detect = Detect.forwardc4 = max(ch[0] // 4, self.nm)self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, self.nm, 1)) for x in ch)def forward(self, x):p = self.proto(x[0])bs = p.shape[0]mc = torch.cat([self.cv4[i](x[i]).view(bs, self.nm, -1) for i in range(self.nl)], 2)  # mask coefficientsx = self.detect(self, x)if self.training:return x, mc, preturn (torch.cat([x, mc], 1), p) if self.export else (torch.cat([x[0], mc], 1), (x[1], mc, p))class Panoptic(Detect):# YOLO Panoptic head for panoptic segmentation modelsdef __init__(self, nc=80, sem_nc=93, nm=32, npr=256, ch=(), inplace=True):super().__init__(nc, ch, inplace)self.sem_nc = sem_ncself.nm = nm  # number of masksself.npr = npr  # number of protosself.proto = Proto(ch[0], self.npr, self.nm)  # protosself.uconv = UConv(ch[0], ch[0]//4, self.sem_nc+self.nc)self.detect = Detect.forwardc4 = max(ch[0] // 4, self.nm)self.cv4 = nn.ModuleList(nn.Sequential(Conv(x, c4, 3), Conv(c4, c4, 3), nn.Conv2d(c4, self.nm, 1)) for x in ch)def forward(self, x):p = self.proto(x[0])s = self.uconv(x[0])bs = p.shape[0]mc = torch.cat([self.cv4[i](x[i]).view(bs, self.nm, -1) for i in range(self.nl)], 2)  # mask coefficientsx = self.detect(self, x)if self.training:return x, mc, p, sreturn (torch.cat([x, mc], 1), p, s) if self.export else (torch.cat([x[0], mc], 1), (x[1], mc, p, s))class BaseModel(nn.Module):# YOLO base modeldef forward(self, x, profile=False, visualize=False):return self._forward_once(x, profile, visualize)  # single-scale inference, traindef _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)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 xdef _profile_one_layer(self, m, x, dt):c = m == self.model[-1]  # is final layer, copy input as inplace fixo = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPst = time_sync()for _ in range(10):m(x.copy() if c else x)dt.append((time_sync() - t) * 100)if m == self.model[0]:LOGGER.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s}  module")LOGGER.info(f'{dt[-1]:10.2f} {o:10.2f} {m.np:10.0f}  {m.type}')if c:LOGGER.info(f"{sum(dt):10.2f} {'-':>10s} {'-':>10s}  Total")def fuse(self):  # fuse model Conv2d() + BatchNorm2d() layersLOGGER.info('Fusing layers... ')for m in self.model.modules():if isinstance(m, (Conv, DWConv)) and hasattr(m, 'bn'):m.conv = fuse_conv_and_bn(m.conv, m.bn)  # update convdelattr(m, 'bn')  # remove batchnormm.forward = m.forward_fuse  # update forwardself.info()return selfdef info(self, verbose=False, img_size=640):  # print model informationmodel_info(self, verbose, img_size)def _apply(self, fn):# Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffersself = super()._apply(fn)m = self.model[-1]  # Detect()if isinstance(m, (Detect, DualDetect, TripleDetect, DDetect, DualDDetect, TripleDDetect, Segment)):m.stride = fn(m.stride)m.anchors = fn(m.anchors)m.strides = fn(m.strides)# m.grid = list(map(fn, m.grid))return selfclass DetectionModel(BaseModel):# YOLO detection modeldef __init__(self, cfg='yolo.yaml', ch=3, nc=None, anchors=None):  # model, input channels, number of classessuper().__init__()if isinstance(cfg, dict):self.yaml = cfg  # model dictelse:  # is *.yamlimport yaml  # for torch hubself.yaml_file = Path(cfg).namewith open(cfg, encoding='ascii', errors='ignore') as f:self.yaml = yaml.safe_load(f)  # model dict# Define modelch = self.yaml['ch'] = self.yaml.get('ch', ch)  # input channelsif nc and nc != self.yaml['nc']:LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")self.yaml['nc'] = nc  # override yaml valueif anchors:LOGGER.info(f'Overriding model.yaml anchors with anchors={anchors}')self.yaml['anchors'] = round(anchors)  # override yaml valueself.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch])  # model, savelistself.names = [str(i) for i in range(self.yaml['nc'])]  # default namesself.inplace = self.yaml.get('inplace', True)# Build strides, anchorsm = self.model[-1]  # Detect()if isinstance(m, (Detect, DDetect, Segment)):s = 256  # 2x min stridem.inplace = self.inplaceforward = lambda x: self.forward(x)[0] if isinstance(m, (Segment)) else self.forward(x)m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))])  # forward# check_anchor_order(m)# m.anchors /= m.stride.view(-1, 1, 1)self.stride = m.stridem.bias_init()  # only run onceif isinstance(m, (DualDetect, TripleDetect, DualDDetect, TripleDDetect)):s = 256  # 2x min stridem.inplace = self.inplace#forward = lambda x: self.forward(x)[0][0] if isinstance(m, (DualSegment)) else self.forward(x)[0]forward = lambda x: self.forward(x)[0]m.stride = torch.tensor([s / x.shape[-2] for x in forward(torch.zeros(1, ch, s, s))])  # forward# check_anchor_order(m)# m.anchors /= m.stride.view(-1, 1, 1)self.stride = m.stridem.bias_init()  # only run once# Init weights, biasesinitialize_weights(self)self.info()LOGGER.info('')def forward(self, x, augment=False, profile=False, visualize=False):if augment:return self._forward_augment(x)  # augmented inference, Nonereturn self._forward_once(x, profile, visualize)  # single-scale inference, traindef _forward_augment(self, x):img_size = x.shape[-2:]  # height, widths = [1, 0.83, 0.67]  # scalesf = [None, 3, None]  # flips (2-ud, 3-lr)y = []  # outputsfor si, fi in zip(s, f):xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max()))yi = self._forward_once(xi)[0]  # forward# cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1])  # saveyi = self._descale_pred(yi, fi, si, img_size)y.append(yi)y = self._clip_augmented(y)  # clip augmented tailsreturn torch.cat(y, 1), None  # augmented inference, traindef _descale_pred(self, p, flips, scale, img_size):# de-scale predictions following augmented inference (inverse operation)if self.inplace:p[..., :4] /= scale  # de-scaleif flips == 2:p[..., 1] = img_size[0] - p[..., 1]  # de-flip udelif flips == 3:p[..., 0] = img_size[1] - p[..., 0]  # de-flip lrelse:x, y, wh = p[..., 0:1] / scale, p[..., 1:2] / scale, p[..., 2:4] / scale  # de-scaleif flips == 2:y = img_size[0] - y  # de-flip udelif flips == 3:x = img_size[1] - x  # de-flip lrp = torch.cat((x, y, wh, p[..., 4:]), -1)return pdef _clip_augmented(self, y):# Clip YOLO augmented inference tailsnl = self.model[-1].nl  # number of detection layers (P3-P5)g = sum(4 ** x for x in range(nl))  # grid pointse = 1  # exclude layer counti = (y[0].shape[1] // g) * sum(4 ** x for x in range(e))  # indicesy[0] = y[0][:, :-i]  # largei = (y[-1].shape[1] // g) * sum(4 ** (nl - 1 - x) for x in range(e))  # indicesy[-1] = y[-1][:, i:]  # smallreturn yModel = DetectionModel  # retain YOLO 'Model' class for backwards compatibilityclass SegmentationModel(DetectionModel):# YOLO segmentation modeldef __init__(self, cfg='yolo-seg.yaml', ch=3, nc=None, anchors=None):super().__init__(cfg, ch, nc, anchors)class ClassificationModel(BaseModel):# YOLO classification modeldef __init__(self, cfg=None, model=None, nc=1000, cutoff=10):  # yaml, model, number of classes, cutoff indexsuper().__init__()self._from_detection_model(model, nc, cutoff) if model is not None else self._from_yaml(cfg)def _from_detection_model(self, model, nc=1000, cutoff=10):# Create a YOLO classification model from a YOLO detection modelif isinstance(model, DetectMultiBackend):model = model.model  # unwrap DetectMultiBackendmodel.model = model.model[:cutoff]  # backbonem = model.model[-1]  # last layerch = m.conv.in_channels if hasattr(m, 'conv') else m.cv1.conv.in_channels  # ch into modulec = Classify(ch, nc)  # Classify()c.i, c.f, c.type = m.i, m.f, 'models.common.Classify'  # index, from, typemodel.model[-1] = c  # replaceself.model = model.modelself.stride = model.strideself.save = []self.nc = ncdef _from_yaml(self, cfg):# Create a YOLO classification model from a *.yaml fileself.model = Nonedef parse_model(d, ch):  # model_dict, input_channels(3)# Parse a YOLO 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)layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch outfor i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, argsm = eval(m) if isinstance(m, str) else m  # eval stringsfor j, a in enumerate(args):with contextlib.suppress(NameError):args[j] = eval(a) if isinstance(a, str) else a  # eval stringsn = n_ = max(round(n * gd), 1) if n > 1 else n  # depth gainif m in {Conv, AConv, ConvTranspose, Bottleneck, SPP, SPPF, DWConv, BottleneckCSP, nn.ConvTranspose2d, DWConvTranspose2d, SPPCSPC, ADown,RepNCSPELAN4, SPPELAN}:c1, c2 = ch[f], args[0]if c2 != no:  # if not outputc2 = make_divisible(c2 * gw, 8)if m in (RepNCSPELAN4, ):args[1] = make_divisible(args[1] * gw, 8)args[2] = make_divisible(args[2] * gw, 8)args = [c1, c2, *args[1:]]if m in {BottleneckCSP, SPPCSPC}: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)elif m is Shortcut:c2 = ch[f[0]]elif m is ReOrg:c2 = ch[f] * 4elif m is CBLinear:c2 = [make_divisible(i * gw, 8) for i in args[0]]c1 = ch[f]args = [c1, c2, *args[1:]]elif m is CBFuse:c2 = ch[f[-1]]# TODO: channel, gw, gdelif m in {Detect, DualDetect, TripleDetect, DDetect, DualDDetect, TripleDDetect, Segment}:args.append([ch[x] for x in f])# if isinstance(args[1], int):  # number of anchors#     args[1] = [list(range(args[1] * 2))] * len(f)if m in {Segment}:args[2] = make_divisible(args[2] * gw, 8)elif m is Contract:c2 = ch[f] * args[0] ** 2elif m is Expand:c2 = ch[f] // args[0] ** 2else:c2 = ch[f]m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args)  # modulet = str(m)[8:-2].replace('__main__.', '')  # module typenp = sum(x.numel() for x in m_.parameters())  # number paramsm_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number paramsLOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # printsave.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 = []ch.append(c2)return nn.Sequential(*layers), sorted(save)if __name__ == '__main__':parser = argparse.ArgumentParser()parser.add_argument('--cfg', type=str, default='yolo.yaml', help='model.yaml')parser.add_argument('--batch-size', type=int, default=1, help='total batch size for all GPUs')parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')parser.add_argument('--profile', action='store_true', help='profile model speed')parser.add_argument('--line-profile', action='store_true', help='profile model speed layer by layer')parser.add_argument('--test', action='store_true', help='test all yolo*.yaml')opt = parser.parse_args()opt.cfg = check_yaml(opt.cfg)  # check YAMLprint_args(vars(opt))device = select_device(opt.device)# Create modelim = torch.rand(opt.batch_size, 3, 640, 640).to(device)model = Model(opt.cfg).to(device)model.eval()# Optionsif opt.line_profile:  # profile layer by layermodel(im, profile=True)elif opt.profile:  # profile forward-backwardresults = profile(input=im, ops=[model], n=3)elif opt.test:  # test all modelsfor cfg in Path(ROOT / 'models').rglob('yolo*.yaml'):try:_ = Model(cfg)except Exception as e:print(f'Error in {cfg}: {e}')else:  # report fused model summarymodel.fuse()