Pytorch里DNN添加batch normalization层；RuntimeError: running_mean should contain 1 elements not 300

Pytorch里nn.BacthNorm1d(num_features)表示添加批标准化。其中num_features不是指的线性层的输出维度！！而是指输入数据的channel维度，一般DNN里肯定为1吧，DNN不会有通道数为3的输入吧，所以:nn.BatchNorm1d(num_features=1)

https://pytorch.org/docs/stable/nn.html#torch.nn.BatchNorm1d

可行的代码如下

self.model=nn.Sequential()
self.model.add_module('fc0',torch.nn.Linear(input_dimension,dim_hidden,bias=True))
self.model.add_module('bn0',nn.BatchNorm1d(num_features=1))
self.model.add_module('act0',activation(act_name))
       
for i in range(1,hidden_layers):
      self.model.add_module(f'fc{i}',nn.Linear(dim_hidden,dim_hidden,bias=True)) 
      self.model.add_module(f'bn{i}',nn.BatchNorm1d(num_features=1))    
      self.model.add_module(f'act{i}',activation(act_name))
      self.model.add_module(f'fc{hidden_layers}',nn.Linear(dim_hidden,output_dimension,bias=True))

注：我的输入数据维度为10，利用Dataloader分batch 取数据，输入数据的处理

features=torch.Tensor(features).view(1,10)