def __init__(self): super(Generator,self).__init__() #first layer of convolution without the residual block self.conv1=nn.Sequuential(nn.Conv2d(3,64,9,1,4), nn.Prelu()) #a serie of residual blocks (8) resBlocks=[] for i in range(8): resBlocks.append(ResBlock(64)) self.resBlocks=nn.Sequuential(*resBlocks) #second convolution after the residual blocks self.conv2=nn.Sequential(nn.Conv2d(64,64,3,1,1),nn.BatchNorm2D(64,0.8)) # upsampling layer with PixelShuffle like stated in the paper self.upsampling=nn.Sequential(nn.Conv2d(64,256,3,1,1), nn.BatchNorm2D(256), nn.PixelShuffle(2), nn.PRelu(), nn.Conv2d(64,256,3,1,1), nn.BatchNorm2D(256), nn.PixelShuffle(2), nn.PRelu()) self.conv3=nn.Sequential(nn.Conv2D(64,3,9,1,4),nn.Tahn())
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(3, 6, 5)
self.pool = nn.MaxPool2d(2, 2)
self.BN1 = nn.BatchNorm2D(6)
self.conv2 = nn.Conv2d(6, 16, 5)
self.BN2 = nn.BatchNorm2D(16)
self.fc1 = nn.Linear(16 * 5 * 5, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def __init__(self,embed_size): super(CNN_Encoder,self).__init__() # Load the pretrained ResNet-152 model and replace the top fc layer resnet = models.resnet(pretrained=True) modules = list[resnet.children()[:,-1]] self.resnet = nn.Sequential(*modules) self.linear = nn.Linear(resnet.fc.in_features,embed_size) self.bn = nn.BatchNorm2D(embed_size,momentum = 0.01)
def __init__(self,
sp1,
sp2,
func,
num_channels,
hidden_size=20000,
bias=True):
super(SparseMNIST_CNN_BN, self).__init__()
# sp1：sp1稀疏比；也可以为k值
# sp2：sp2稀疏比
# func：k-WTA/ReLU激活函数
# num_channels：每一层的通道数
# hidden_size：隐藏层神经元数量，默认为20000
# bias：是否求偏差，默认为True
self.cnn = nn.Sequential(
nn.Conv2d(1, num_channels[0], 3, padding=1, bias=bias),
nn.BatchNorm2D(channels[0]),
activation_function.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[0],
num_channels[1],
3,
padding=1,
stride=2,
bias=bias),
nn.BatchNorm2D(channels[1]),
activation_function.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[1],
num_channels[2],
3,
padding=1,
bias=bias),
nn.BatchNorm2D(channels[2]),
activation_function.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[2],
num_channels[3],
3,
padding=1,
stride=2,
bias=bias),
nn.BatchNorm2D(channels[3]),
activation_function.sparse_func_dict[func](sp1),
Flatten(),
nn.Linear(7 * 7 * num_channels[3], hidden_size),
activation_function.Sparsify1D(sp2),
# nn.Linear(hidden_size, hidden_size), activation_function.Sparsify1D(sp),
nn.Linear(hidden_size, 10))
def __init__(self,
sp1,
sp2,
func,
num_channels,
hidden_size=20000,
bias=True):
super(SparseMNIST_CNN_BN, self).__init__()
self.cnn = nn.Sequential(
nn.Conv2d(1, num_channels[0], 3, padding=1, bias=bias),
nn.BatchNorm2D(channels[0]),
models.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[0],
num_channels[1],
3,
padding=1,
stride=2,
bias=bias),
nn.BatchNorm2D(channels[1]),
models.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[1],
num_channels[2],
3,
padding=1,
bias=bias),
nn.BatchNorm2D(channels[2]),
models.sparse_func_dict[func](sp1),
nn.Conv2d(num_channels[2],
num_channels[3],
3,
padding=1,
stride=2,
bias=bias),
nn.BatchNorm2D(channels[3]),
models.sparse_func_dict[func](sp1),
Flatten(),
nn.Linear(7 * 7 * num_channels[3], hidden_size),
models.Sparsify1D(sp2),
# nn.Linear(hidden_size, hidden_size), models.Sparsify1D(sp),
nn.Linear(hidden_size, 10))
def __init__(self):
super(simpleModel, self).__init__()
self.conv = nn.Conv2D(in_channels=3,
out_channels=16,
kernel_size=3,
padding=1,
strides=1,
bias=False)
self.batchnorm = nn.BatchNorm2D(16, affine=True)
self.relu = nn.Activation('relu')
self.avg_pool = AvgPool2D(8)
self.flatten = Flatten()
self.fc = nn.Linear(256,
10) #torch에는 softmax가 없고 대신 crossetnrotpyloss를 쓴다
def __init__(self, in_channels,out_channels, **kwargs):
super().__init__(self)
self.conv1 = nn.Conv2d(in_channels, out_channels,**kwargs)
self.bn1 =nn.BatchNorm2D(out_channels)
self.relu = nn.Relu()