nn.Linear nn.Conv2d nn.BatchNorm2d

conv,BN,Linear
conv:https://blog.csdn.net/Strive_For_Future/article/details/83240232
1)conv2d.weight     shape=[输出channels,输入channels,kernel_size,kernel_size]

2)conv2d.bias   shape=[输出channels]
BN:https://www.cnblogs.com/tingtin/p/12523701.html
尺寸：输入输出一样
m = nn.BatchNorm2d(2,affine=True) #2表示输出通道数，affine=True表示权重w和偏重b将被使用学习
m.weight:tensor([1., 1.], requires_grad=True)
m.bias:tensor([0., 0.], requires_grad=True)#w,b都是大小维输出通道数的向量
Linear:https://www.cnblogs.com/tingtin/p/12425849.html
nn.Linear()用于设置全连接层，输入输出均为二维张量，形状为[batch_size, size]
def __init__(self, in_features, out_features, bias=True):
        super(Linear, self).__init__()
        self.in_features = in_features
        self.out_features = out_features
        self.weight = Parameter(torch.Tensor(out_features, in_features))
        if bias:
            self.bias = Parameter(torch.Tensor(out_features))
        else:
            self.register_parameter('bias', None)
        self.reset_parameters()
nn.Linear nn.Conv2d nn.BatchNorm2d

推荐阅读
