Pytorch and Np / pd notes

1.CSV turn Matix

import pandas as pd
dat = pd.read.csv('path')
dat_matrix = df.values

2.Matrix sliced

dat_m_0 = dat_matrix[:,1]  #截取矩阵中第二列所有数据
dat_m_1 = dat_matrix[:,1:5]  #截取矩阵中第一至四列所有数据
dat_m_2 = dat_matrix[1,:]  #截取矩阵中第二行所有数据
dat_m_3 = dat_matrix[1:5,:]  #截取矩阵中第二至四行所有数据

3.Matrix turn torch

dat_array = np.array(dat_matrix)
dat_torch = torch.from_numpy(dat_array)

4.torch data type change

dat_t_float = dat_torch.float()
dat_t_long = dat_torch.long()
dat_t_int32 = dat_torch.int()

5. With regard to the number of neurons

• 1. The number of neurons required dimension corresponding to the number of the input data
• 2. The output neurons corresponding to the category

6. For the loss function

• 1.CrossEntropy (), y data required training must be long type
• 2.MSELoss (), y data requires training to be a float

Reproduced in: https: //www.jianshu.com/p/1412a9995f01