Python RuntimeError:Generator在加载状态下的错误：使用Pyrotch的权重和偏差大小不匹配
我正在训练一个3D-GAN来生成MRI体积。我将我的模型定义如下：Python RuntimeError:Generator在加载状态下的错误：使用Pyrotch的权重和偏差大小不匹配,python,pytorch,generative-adversarial-network,Python,Pytorch,Generative Adversarial Network,我正在训练一个3D-GAN来生成MRI体积。我将我的模型定义如下： ###### Definition of the generator ###### class Generator(nn.Module): def __init__(self, ngpu): #super() makes Generator a subclass of nn.Module, so that it inherites all the methods of nn.Module super(Gen
###### Definition of the generator ######
class Generator(nn.Module):
def __init__(self, ngpu):
#super() makes Generator a subclass of nn.Module, so that it inherites all the methods of nn.Module
super(Generator, self).__init__()
self.ngpu = ngpu
#we can use Sequential() since the output of one layer is the input of the next one
self.main = nn.Sequential(
# input is latent vector z, going into a convolution
nn.ConvTranspose3d(nz, ngf * 8, 4, stride=2, padding=0, bias=True), # try to put kernel = (batch_size,4,4,4,512)
nn.BatchNorm3d(ngf * 8),
nn.ReLU(True), #True means that it does the operation inplace, default is False
nn.ConvTranspose3d(ngf * 8, ngf * 4, 4, stride=2, padding=1, bias=True), # try to put kernel = (batch_size,8,8,8,256)
nn.BatchNorm3d(ngf * 4),
nn.ReLU(True),
nn.ConvTranspose3d(ngf * 4, ngf * 2, 4, stride=2, padding=1, bias=True), # try to put kernel = (batch_size,16,16,16,128)
nn.BatchNorm3d(ngf * 2),
nn.ReLU(True),
nn.ConvTranspose3d( ngf * 2, ngf, 4, stride=2, padding=1, bias=True), # try to put kernel = (batch_size,32,32,32,64)
nn.BatchNorm3d(ngf),
nn.ReLU(True),
nn.ConvTranspose3d(ngf, nc, 4, stride=2, padding=1, bias=True), # try to put kernel = (batch_size,64,64,64,1)
nn.Sigmoid()
)
def forward(self, x):
return self.main(x)
###### Definition of the Discriminator ######
class Discriminator(nn.Module):
def __init__(self, ngpu):
super(Discriminator, self).__init__()
self.ngpu = ngpu
self.main = nn.Sequential(
nn.Conv3d(nc, ndf, 4, stride=2, padding=1, bias=True),
nn.BatchNorm3d(ndf),
nn.LeakyReLU(leak_value, inplace=True),
nn.Conv3d(ndf, ndf * 2, 4, stride=2, padding=1, bias=True),
nn.BatchNorm3d(ndf * 2),
nn.LeakyReLU(leak_value, inplace=True),
nn.Conv3d(ndf * 2, ndf * 4, 4, stride=2, padding=1, bias=True),
nn.BatchNorm3d(ndf * 4),
nn.LeakyReLU(leak_value, inplace=True),
nn.Conv3d(ndf * 4, ndf * 8, 4, stride=2, padding=1, bias=True),
nn.BatchNorm3d(ndf * 8),
nn.LeakyReLU(leak_value, inplace=True),
nn.Conv3d(ndf * 8, nc, 4, stride=1, padding=0, bias=True),
nn.Sigmoid()
)
def forward(self, x):
return self.main(x)
提前谢谢 您加载的模型与目标模型不相同，因此会出现错误，提示大小、图层不匹配，请再次检查代码，否则保存的模型可能无法正确保存您确定两次都使用相同的代码从类创建模型吗？因为误差表明评估时间模型与训练时间模型不同。您可能需要向任何人显示更多代码才能识别问题。这是正确的。如果在训练和评估之间更改模型中的权重大小和偏差，则可能会发生这种情况。但是，如果这是需要的行为，则在加载
state\u dictstrict=False