YOLOv5代码解读[02] models/yolov5l.yaml文件解析

news/2024/4/23 2:55:43

文章目录

  • YOLOv5代码解读[02] models/yolov5l.yaml文件解析
    • yolov5l.yaml文件
    • 检测头1--->耦合头
    • 检测头2--->解耦头
    • 检测头3--->ASFF检测头
    • Model类解析
    • parse_model函数

YOLOv5代码解读[02] models/yolov5l.yaml文件解析

yolov5l.yaml文件

在这里插入图片描述

# YOLOv5 🚀 by Ultralytics, GPL-3.0 license# Parameters
nc: 27  # number of classes
depth_multiple: 1.0  # model depth multiple
width_multiple: 1.0  # 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, Conv, [64, 6, 2, 2]],  # 0-P1/2[-1, 1, Conv, [128, 3, 2]],  # 1-P2/4[-1, 3, C3, [128]],[-1, 1, Conv, [256, 3, 2]],  # 3-P3/8[-1, 6, C3, [256]],[-1, 1, Conv, [512, 3, 2]],  # 5-P4/16[-1, 9, C3, [512]],[-1, 1, Conv, [1024, 3, 2]],  # 7-P5/32[-1, 3, C3, [1024]],[-1, 1, SPPF, [1024, 5]],  # 9]# YOLOv5 v6.0 head
head:[[-1, 1, Conv, [512, 1, 1]],[-1, 1, nn.Upsample, [None, 2, 'nearest']],[[-1, 6], 1, Concat, [1]],  # cat backbone P4[-1, 3, C3, [512, False]],  # 13[-1, 1, Conv, [256, 1, 1]],[-1, 1, nn.Upsample, [None, 2, 'nearest']],[[-1, 4], 1, Concat, [1]],  # cat backbone P3[-1, 3, C3, [256, False]],  # 17 (P3/8-small)[-1, 1, Conv, [256, 3, 2]],[[-1, 14], 1, Concat, [1]],  # cat head P4[-1, 3, C3, [512, False]],  # 20 (P4/16-medium)[-1, 1, Conv, [512, 3, 2]],[[-1, 10], 1, Concat, [1]],  # cat head P5[-1, 3, C3, [1024, False]],  # 23 (P5/32-large)[[17, 20, 23], 1, Detect, [nc, anchors, False]],  # Detect(P3, P4, P5)]

检测头1—>耦合头

class Detect(nn.Module):stride = None  onnx_dynamic = Falseexport = Falsedef __init__(self, nc=80, anchors=(), Decoupled=False, ch=(), inplace=True):  super().__init__()# 是否解耦头self.decoupled = Decoupled# 类别数目self.nc = nc  # 每个anchor输出维度 self.no = nc + 5  # 检测层的输出数量(不同尺度个数) self.nl = len(anchors)  # 每个尺度特征图的anchor数量self.na = len(anchors[0]) // 2  # 初始化步长init gridself.grid = [torch.zeros(1)] * self.nl    # 初始化anchor gridself.anchor_grid = [torch.zeros(1)] * self.nl  # self.register_buffer("a", torch.ones(2,3))  # register_buffer的作用是将torch.ones(2,3)这个tensor注册到模型的buffers()属性中,并命名为a,# 这代表a对应的是一个持久态,不会有梯度传播给它,但是能被模型的state_dict记录下来,可以理解为模型的常数。self.register_buffer('anchors', torch.tensor(anchors).float().view(self.nl, -1, 2))  # (3,3,2) == (nl,na,2)# 检测头head输出卷积# 如果是解耦头if self.decoupled:self.m = nn.ModuleList(DecoupledHead(x, self.nc, anchors) for x in ch) # 如果是耦合头else:self.m = nn.ModuleList(nn.Conv2d(x, self.no*self.na, 1) for x in ch) # use in-place ops (e.g. slice assignment)self.inplace = inplace  def forward(self, x):# inference outputz = []# 对于每个尺度的特征图来说for i in range(self.nl):# conv# P3: [1, 128, 80, 80]->[1, 3*(nc+5), 80, 80]# P4: [1, 256, 40, 40]->[1, 3*(nc+5), 40, 40]# P5: [1, 512, 20, 20]->[1, 3*(nc+5), 20, 20]x[i] = self.m[i](x[i])# 以coco数据集为例,x(bs,255,20,20) -> x(bs,3,20,20,85)   (x,y,w,h,c,c1,c2,.........)bs, _, ny, nx = x[i].shapex[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()# 推断过程inferenceif not self.training:# self.grid: [tensor([0.]), tensor([0.]), tensor([0.])]if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)y = x[i].sigmoid()if self.inplace:# 中心点xy 网格gridy[..., 0:2] = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]# 长宽wh  锚anchor_gridy[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]else:xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]y = torch.cat((xy, wh, y[..., 4:]), -1)z.append(y.view(bs, -1, self.no))return x if self.training else (torch.cat(z, 1),) if self.export else (torch.cat(z, 1), x)# # 转成caffe时候的代码# def forward(self, x):#     # inference output#     z = []#     # 对于每个尺度的特征图来说#     for i in range(self.nl):#         # conv#         # P3: [1, 128, 80, 80]->[1, 3*(nc+5), 80, 80]#         # P4: [1, 256, 40, 40]->[1, 3*(nc+5), 40, 40]#         # P5: [1, 512, 20, 20]->[1, 3*(nc+5), 20, 20]#         x[i] = self.m[i](x[i])#         # y = x[i]#         y = x[i].sigmoid()#         z.append(y)#     return zdef _make_grid(self, nx=20, ny=20, i=0, torch_1_10=check_version(torch.__version__, '1.10.0')):d = self.anchors[i].devicet = self.anchors[i].dtypey, x = torch.arange(ny, device=d, dtype=t), torch.arange(nx, device=d, dtype=t)# torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibilityif torch_1_10:yv, xv = torch.meshgrid(y, x, indexing='ij')else:yv, xv = torch.meshgrid(y, x)# 网格grid (x, y)# x[i] --> (bs,3,ny,nx,85)# grid --> (1,3,ny,nx,2)grid = torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2))# 锚anchor (w, h)# x[i] --> (bs,3,ny,nx,85)# anchor_grid --> (1,3,ny,nx,2)# self.stride: tensor([ 8., 16., 32.])anchor_grid = (self.anchors[i].clone() * self.stride[i]).view((1, self.na, 1, 1, 2)).expand((1, self.na, ny, nx, 2))return grid, anchor_grid

检测头2—>解耦头

class DecoupledHead(nn.Module):def __init__(self, ch=256, nc=80, anchors=()):super().__init__()# 类别个数self.nc = nc# 检测层的数量self.nl = len(anchors)# 每一层anchor个数self.na = len(anchors[0]) // 2self.merge = Conv(ch, 128 , 1, 1)  # 默认256self.cls_convs1 = Conv(128, 64, 3, 1, 1)self.cls_convs2 = Conv(64, 64, 3, 1, 1)self.reg_convs1 = Conv(128, 64, 3, 1, 1)self.reg_convs2 = Conv(64, 64, 3, 1, 1)self.cls_preds = nn.Conv2d(64 , self.nc*self.na, 1)self.reg_preds = nn.Conv2d(64 , 4*self.na, 1)self.obj_preds = nn.Conv2d(64 , 1*self.na, 1)def forward(self, x):x = self.merge(x)x1 = self.cls_convs1(x)x1 = self.cls_convs2(x1)x1 = self.cls_preds(x1)x2 = self.reg_convs1(x)x2 = self.reg_convs2(x2)x21 = self.reg_preds(x2)x22 = self.obj_preds(x2)out = torch.cat([x21, x22, x1], 1)return out

检测头3—>ASFF检测头

class ASFF_Detect(nn.Module):  stride = None  onnx_dynamic = False   def __init__(self, nc=80, anchors=(), ch=(), multiplier=0.5, rfb=False, inplace=True):  super().__init__()# 类别数目self.nc = nc  # 每个anchor输出维度self.no = nc + 5  # 检测层的输出数量(不同尺度个数) self.nl = len(anchors) # 每个尺度特征图的anchor数量self.na = len(anchors[0]) // 2  # 初始化步长init gridself.grid = [torch.zeros(1)] * self.nl  # init anchor gridself.anchor_grid = [torch.zeros(1)] * self.nl# self.register_buffer("a", torch.ones(2,3))  # register_buffer的作用是将torch.ones(2,3)这个tensor注册到模型的buffers()属性中,并命名为a,# 这代表a对应的是一个持久态,不会有梯度传播给它,但是能被模型的state_dict记录下来,可以理解为模型的常数。self.register_buffer('anchors', torch.tensor(anchors).float().view(self.nl, -1, 2))  # (3,3,2) == (nl,na,2)# ASFF模块self.l0_fusion = ASFFV5(level=0, multiplier=multiplier, rfb=rfb)self.l1_fusion = ASFFV5(level=1, multiplier=multiplier, rfb=rfb)self.l2_fusion = ASFFV5(level=2, multiplier=multiplier, rfb=rfb)# 检测头head输出卷积self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch)  # use in-place ops (e.g. slice assignment)self.inplace = inplace  def forward(self, x):# inference outputz = []  result = []result.append(self.l2_fusion(x))result.append(self.l1_fusion(x))result.append(self.l0_fusion(x))x = result    # 对于每个尺度的特征图来说for i in range(self.nl):# conv # P3: [1, 128, 80, 80]->[1, 3*(nc+5), 80, 80]# P4: [1, 256, 40, 40]->[1, 3*(nc+5), 40, 40]# P5: [1, 512, 20, 20]->[1, 3*(nc+5), 20, 20]x[i] = self.m[i](x[i])  # 以coco数据集为例,x(bs,255,20,20) -> x(bs,3,20,20,85)   (x,y,w,h,c,c1,c2,.........)bs, _, ny, nx = x[i].shape  x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()# 推断过程inference if not self.training:  # self.grid: [tensor([0.]), tensor([0.]), tensor([0.])]if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i)y = x[i].sigmoid()# 这块xy的计算存在大量疑惑?????????????????????????if self.inplace:# 中心点xy 网格gridy[..., 0:2] = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  # 长宽wh  锚anchor_gridy[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] else:  # for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i]  wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  y = torch.cat((xy, wh, y[..., 4:]), -1)z.append(y.view(bs, -1, self.no))return x if self.training else (torch.cat(z, 1), x)def _make_grid(self, nx=20, ny=20, i=0):d = self.anchors[i].deviceif check_version(torch.__version__, '1.10.0'):  # torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibilityyv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)], indexing='ij')else:yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)])# 网格grid (x, y)# x[i] --> (bs,3,ny,nx,85)# grid --> (1,3,ny,nx,2)grid = torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2)).float()# 锚anchor (w, h)# x[i] --> (bs,3,ny,nx,85)# anchor_grid --> (1,3,ny,nx,2)# self.stride: tensor([ 8., 16., 32.])anchor_grid = (self.anchors[i].clone() * self.stride[i]).view((1, self.na, 1, 1, 2)).expand((1, self.na, ny, nx, 2)).float()return grid, anchor_grid

Model类解析

class Model(nn.Module):def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None):  super().__init__()# 字典dict类型if isinstance(cfg, dict):self.yaml = cfg  # yaml文件else: self.yaml_file = Path(cfg).name# 用ascii编码,忽略错误的形式打开文件cfgwith open(cfg, encoding='ascii', errors='ignore') as f:self.yaml = yaml.safe_load(f)  # 输入通道ch = self.yaml['ch'] = self.yaml.get('ch', ch)  # 重写yaml文件中的ncif nc and nc != self.yaml['nc']:LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}")self.yaml['nc'] = nc  # 重写yaml文件中的anchors if anchors:LOGGER.info(f'Overriding model.yaml anchors with anchors={anchors}')self.yaml['anchors'] = round(anchors)  # 根据yaml文件的model_dict解析模型self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # 默认类别名字 从0到nc-1self.names = [str(i) for i in range(self.yaml['nc'])] self.inplace = self.yaml.get('inplace', True)# 设置Detect()中的inplace, stride, anchorsm = self.model[-1]  if isinstance(m, Detect) or isinstance(m, ASFF_Detect):s = 256m.inplace = self.inplace# 根据前向传播forward 计算步长stridem.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))])# 把anchors放缩到了3个不同的尺度# 这块的形状为什么这样变化??????m.anchors /= m.stride.view(-1, 1, 1)# 根据YOLOv5 Detect()模块m的步幅顺序检查给定锚框顺序,必要时进行纠正。check_anchor_order(m)self.stride = m.strideif m.decoupled:LOGGER.info('decoupled done')pass else:self._initialize_biases()  # only run once  # 初始化权重weights和偏置biasesinitialize_weights(self)self.info()LOGGER.info('')def forward(self, x, augment=False, profile=False, visualize=False):# 推断时增强augmented inferenceif augment:return self._forward_augment(x)  # 单尺度推断single-scale inference 或者训练trainreturn self._forward_once(x, profile, visualize)  def _forward_augment(self, x):# height, widthimg_size = x.shape[-2:]  s = [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 _forward_once(self, x, profile=False, visualize=False):y, dt = [], []  for m in self.model:# 输入不是来自于上一个层的输出if m.f != -1:  x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]if profile:self._profile_one_layer(m, x, dt)# 计算输出x = m(x)y.append(x if m.i in self.save else None) # 特征可视化if visualize:feature_visualization(x, m.type, m.i, save_dir=visualize)return xdef _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 YOLOv5 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 ydef _profile_one_layer(self, m, x, dt):c = isinstance(m, Detect) or isinstance(m, ASFF_Detect) # 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 _initialize_biases(self, cf=None):  # initialize biases into Detect(), cf is class frequency# https://arxiv.org/abs/1708.02002 section 3.3# cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.m = self.model[-1]  # mi--> Conv2d(128, 255, kernel_size=(1, 1), stride=(1, 1)) # s --> tensor(8.)for mi, s in zip(m.m, m.stride):  # conv.bias(255) to (3,85)b = mi.bias.view(m.na, -1)  b.data[:, 4] += math.log(8 / (640 / s) ** 2)  # obj (8 objects per 640 image)b.data[:, 5:] += math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum())  # clsmi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)def _print_biases(self):m = self.model[-1]  for mi in m.m:  b = mi.bias.detach().view(m.na, -1).T  LOGGER.info(('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))def _print_weights(self):for m in self.model.modules():if type(m) is Bottleneck:LOGGER.info('%10.3g' % (m.w.detach().sigmoid() * 2))  # shortcut weightsdef 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):  # 打印模型信息model_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) or isinstance(m, ASFF_Detect) or isinstance(m, Decoupled_Detect):m.stride = fn(m.stride)m.grid = list(map(fn, m.grid))if isinstance(m.anchor_grid, list):m.anchor_grid = list(map(fn, m.anchor_grid))return self

parse_model函数

def parse_model(d, ch):  # model_dict, input_channels(3)LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10}  {'module':<40}{'arguments':<30}")# nc:类别数; gd:'depth_multiple'; gw:'width_multiple'anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']# anchor数目, 每层为3na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors # 每层的输出,na*(classes+5)no = na * (nc + 5)  # layers, savelist, ch_outlayers, save, c2 = [], [], ch[-1] # from, number, module, args# 以[-1, 1, Conv, [64, 6, 2, 2]为例, ch=[3], f=-1, n=1, m=Conv, args=[64, 6, 2, 2]#   [-1, 1, Conv, [128, 3, 2]#   [-1, 3, C3, [128]]#   [-1, 1, SPPF, [1024, 5]]#   [-1, 1, nn.Upsample, [None, 2, 'nearest']]#   [[-1, 6], 1, Concat, [1]]#   [-1, 3, C3, [512, False]]for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):# 把strings转为本身的类型m = eval(m) if isinstance(m, str) else m  for j, a in enumerate(args):try:# 列表形式args[j] = eval(a) if isinstance(a, str) else a  except NameError:pass# depth_gain 深度缩放因子n = n_ = max(round(n*gd), 1) if n > 1 else n # 对于不同类型的卷积模块   if m in [Conv, DWConv,  CrossConv, GhostConv, Bottleneck, GhostBottleneck,BottleneckCSP, MobileBottleneck, SPP, SPPF, MixConv2d, Focus,InvertedResidual, ConvBNReLU, C3, C3TR, C3SPP, C3Ghost, CoordAtt,CoordAttv2, OSA_Stage]:# i=0, c1=3,  c2=64;  # i=1, c1=32, c2=128;  # i=2, c1=64, c2=128;# c1输入通道;c2输出通道;c1, c2 = ch[f], args[0]# width_gain 宽度缩放因子# 说明不是输出if c2 != no:  # 输出通道数必须为8的倍数c2 = make_divisible(c2*gw, 8)# i=0, [3,  32, 6, 2, 2]# i=1, [32, 64, 3, 2]# i=2, [64, 64]args = [c1, c2, *args[1:]]# 堆叠次数number of repeats# 注意网络设计理念:stage ---> block ---> layerif m in [BottleneckCSP, C3, C3TR, C3Ghost]:args.insert(2, n)  n = 1elif m is nn.BatchNorm2d:args = [ch[f]]elif m is Concat:c2 = sum(ch[x] for x in f)elif m is Detect:args.append([ch[x] for x in f])if isinstance(args[1], int):  # number of anchorsargs[1] = [list(range(args[1] * 2))] * len(f)elif m is ASFF_Detect :args.append([ch[x] for x in f])if isinstance(args[1], int):  # number of anchorsargs[1] = [list(range(args[1] * 2))] * len(f) elif m is Contract:c2 = ch[f] * args[0] ** 2elif m is Expand:c2 = ch[f] // args[0] ** 2elif m is ConvNeXt_Block:c2 = args[0]args = args[1:]else:c2 = ch[f]# module# Conv(3, 32, 6, 2, 2]m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args) # m ===> <class 'models.common.Conv'># str(m)[8:-2] ===> models.common.Convt = str(m)[8:-2].replace('__main__.', '')  # 参数(parameters)/模型参数, 由模型通过学习得到的变量,比如权重和偏置.# m_.parameters(): <generator object Module.parameters at 0x7fcf4c2059d0>np = sum(x.numel() for x in m_.parameters()) # attach index, 'from' index, type, number paramsm_.i, m_.f, m_.type, m_.np = i, f, t, np  LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f}  {t:<40}{str(args):<30}')  # savelist  [6, 4, 14, 10, 17, 20, 23]save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # layers列表layers.append(m_)if i == 0:ch = []# ch列表ch.append(c2)return nn.Sequential(*layers), sorted(save)

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处:http://www.cpky.cn/p/8522.html

如若内容造成侵权/违法违规/事实不符,请联系我们进行投诉反馈,一经查实,立即删除!

相关文章

找座位 - 华为OD统一考试(C卷)

OD统一考试(C卷) 分值: 100分 题解: Java / Python / C++ 题目描述 在一个大型体育场内举办了一场大型活动,由于疫情防控的需要,要求每位观众的必须间隔至少一个空位才允许落座。 现在给出一排观众座位分布图,座位中存在已落座的观众,请计算出,在不移动现有观众座位…

js设计模式:依赖注入模式

作用: 在对象外部完成两个对象的注入绑定等操作 这样可以将代码解耦,方便维护和扩展 vue中使用use注册其他插件就是在外部创建依赖关系的 示例: class App{constructor(appName,appFun){this.appName appNamethis.appFun appFun}}class Phone{constructor(app) {this.nam…

基于SpringBoot的产业园区智慧公寓管理系统

文章目录 项目介绍主要功能截图&#xff1a;部分代码展示设计总结项目获取方式 &#x1f345; 作者主页&#xff1a;超级无敌暴龙战士塔塔开 &#x1f345; 简介&#xff1a;Java领域优质创作者&#x1f3c6;、 简历模板、学习资料、面试题库【关注我&#xff0c;都给你】 &…

如何使用Docker和Nginx部署Web应用

Hello大家好&#xff01;我是咕噜铁蛋今天我要和大家分享一个非常实用的技术教程——如何使用Docker和Nginx来部署Web应用。随着互联网的发展&#xff0c;Web应用越来越普及&#xff0c;而使用Docker和Nginx来部署Web应用已经成为了一种流行的做法。接下来&#xff0c;让我带领…

Linux搭建FISCO BCOS的第一个区块链网络

一、前言 FISCO BCOS是由金融区块链合作联盟&#xff08;深圳&#xff09;与微众银行共同发起的开源区块链项目&#xff0c;支持多链多账本&#xff0c;满足金融行业复杂业务需求。本文将介绍如何在Ubuntu操作系统上使用Linux命令搭建FISCO BCOS的第一个区块链网络。 目录 一…

机器学习——正规方程

正规方程的基本介绍 之前我们使用梯度下降算法求代价函数J(θ)的最小值&#xff0c;而梯度下降算法是通过一步步不断地迭代来收敛到全局最小值&#xff0c;如下 而正规方程则是另一种求解J(θ)最小值的方法&#xff0c;并且正规方程不需要通过迭代&#xff0c;而是一次性得到θ…