Record the errors encountered in the pytorch training model once

1. Missing key(s) in state_dict: "first_conv.bias", 
   1. Solve settings ：model.load_state_dict(checkpoint,strict=False)
2.  TypeError: 'Generator' object is not subscriptable
   1. solve ： Unify saving and loading （ For example, the model is saved , Just load the whole model and then load_state, If you save a super parameter , Then put it directly into the model , no need load_state 了 ）
3.   Report errors ：ValueError: The length of the input vector x must be greater than padlen, which is 18.
   1. solve ：y = signal.filtfilt(b, a, x,axis=0) 
4. Report errors ：ValueError: operands could not be broadcast together with shapes (323596,2) (323596,)
   1. reason ： Output to view the data in the original warehouse x The dimension of is one-dimensional （ Such as (96161,)）, The dimension of the newly added data is two-dimensional （ Such as (321199, 2),(549108, 2)）
   2. Error resolution ：x_tmp=[token for st in x for token in st] This will turn numpy Turned into list, There is no shape The attribute is
   3. Correct the problem ： Yes shape The attributes are array Can be used directly flatten Turn it into 1 Dimensional , If you need to change to another dimension, you can use reshape
5. Expected weight to be a vector of size equal to the number of channels in input, but got weight of shape [512] and input of shape [512, 7622]
   1. solve ： take input To transpose