遇到 TypeError: empty(): argument 'size' must be tuple of SymInts, but found element of type str at pos 3

Question

我遇到了 TypeError: empty(): argument 'size' must be tuple of SymInts, but found element of type str at pos 3

我是这个深度学习领域的新手，只是想将示例 DNN 代码修改为 FCN 代码

所以基本上结构是由其他人编写的，当我尝试修改一些输入参数时

与

我想知道为什么

有人可以给我一些建议吗？

class FCN(nn.Module):
    def __init__(self, args):
        super(FCN,self).__init__()

        model_name = args.model_name
        self.mdl_attrib = args.model_name.split('-')[1]
        
        Input_Size = int(model_name.split('-I_')[1].split('-')[0].split('_')[0])
        

        channel_size_list = list(map(int,re.split(r'[-_]',model_name.split('-M_')[1].split('-O_')[0])[::2]))
        

        #a = type(channel_size_list[0])
        kernel_size_list = re.split(r'[-_]',model_name.split('-M_')[1].split('-O_')[0])[1::2]
        
        Outpt_Size = int(model_name.split('-O')[1].split('-')[0].split('_')[1])

        
        self.NumLyrs = len(channel_size_list) + 2
        
        self.UsedCmpx = False
        if args.use_complex:
            Dtype = torch.cfloat
            self.UsedCmpx = True
        else:
            Dtype = torch.float  

        self.Conv2dLayerList = nn.ModuleList()
        self.ActvationLrList = nn.ModuleList()
        self.InstNormLyrList = nn.ModuleList()


        if self.NumLyrs == 2:
            self.Conv2dLayerList.append(nn.Conv2d(Input_Size, Outpt_Size, dtype = Dtype))
            
        for LyInd in range(self.NumLyrs - 1):
            if LyInd == self.NumLyrs - 2:
                self.Conv2dLayerList.append(nn.Conv2d(channel_size_list[-1], Outpt_Size, kernel_size_list[LyInd],dtype = Dtype))
                if self.mdl_attrib == 'Cls':
                    self.ActvationLrList.append(nn.Sigmoid())        
            elif LyInd == 0:
                self.Conv2dLayerList.append(nn.Conv2d(Input_Size, channel_size_list[LyInd], kernel_size_list[LyInd], dtype = Dtype))
                self.ActvationLrList.append(nn.CELU())
                self.InstNormLyrList.append(nn.InstanceNorm2d(channel_size_list[LyInd]))
            else:
                self.Conv2dLayerList.append(nn.Conv2d(channel_size_list[LyInd-1],channel_size_list[LyInd], kernel_size_list[LyInd], dtype = Dtype))
                self.ActvationLrList.append(nn.CELU())
                self.InstNormLyrList.append(nn.InstanceNorm2d(channel_size_list[LyInd]))
            
    def forward(self, Inp):
    
        if self.NumLyrs == 2:
            LyrOut = self.Conv2dLayerList[0](Inp)
            if self.mdl_attrib == 'Cls':                
                if self.UsedCmpx:
                    RLyrOut = self.ActvationLrList[0](torch.real(LyrOut))
                    ILyrOut = self.ActvationLrList[0](torch.imag(LyrOut))
                    LyrOut = RLyrOut + 1j * ILyrOut
                else:
                    LyrOut = self.ActvationLrList[0](LyrOut)
            return LyrOut
        else:        
            for lyInd in range(self.NumLyrs - 1):
                if lyInd == self.NumLyrs - 2:
                    LyrOut = self.Conv2dLayerList[lyInd](LyrOut)
                    if self.mdl_attrib == 'Cls':                
                        if self.UsedCmpx:
                            RLyrOut = self.ActvationLrList[lyInd](torch.real(LyrOut))
                            ILyrOut = self.ActvationLrList[lyInd](torch.imag(LyrOut))
                            LyrOut = RLyrOut + 1j * ILyrOut
                        else:
                            LyrOut = self.ActvationLrList[lyInd](LyrOut)
                    return LyrOut
                elif lyInd == 0:
                    LyrOut = self.Conv2dLayerList[lyInd](Inp)
                    if self.UsedCmpx:
                        RLyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](torch.real(LyrOut)).permute(0,2,1)).permute(0,2,1)
                        ILyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](torch.imag(LyrOut)).permute(0,2,1)).permute(0,2,1)
                        LyrOut = RLyrOut + 1j * ILyrOut
                    else:
                        LyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](LyrOut))
                else:
                    LyrOut = self.Conv2dLayerList[lyInd](LyrOut)
                    if self.UsedCmpx:
                        RLyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](torch.real(LyrOut)).permute(0,2,1)).permute(0,2,1)
                        ILyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](torch.imag(LyrOut)).permute(0,2,1)).permute(0,2,1)
                        LyrOut = RLyrOut + 1j * ILyrOut
                    else:
                        LyrOut = self.InstNormLyrList[lyInd](self.ActvationLrList[lyInd](LyrOut))

尝试在 linux 系统上运行，结果发生了