Several ways to store and read variables in python

1. numpy library operations

• numpy.save(): Arrays are saved in uncompressed raw binary format in files with a .npy extension.
• numpy.load(): Read the data of the .npy file and directly convert it to a numpy array

☀☀☀<<example>>☀☀☀

>>> import numpy as np
>>> a = np.arange(24).reshape(2,3,4)
>>> a
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],
 
       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]]])
>>> np.save("D:/aa/npp.npy", a)
>>> b = np.load("D:/aa/npp.npy")
>>> b
array([[[ 0,  1,  2,  3],
        [ 4,  5,  6,  7],
        [ 8,  9, 10, 11]],
 
       [[12, 13, 14, 15],
        [16, 17, 18, 19],
        [20, 21, 22, 23]]])

If the error Object arrays cannot be loaded when allow_pickle=False is reported , it needs to be changed as follows:

np.load(path, allow_pickle=True)

2. pandas library operation

• to_pickle: This method is to directly store the dataframe variable data as a local file, and there is no requirement for the file extension
• read_pickle: This method is to read the variables stored locally and convert them into dataframe files

☀☀☀<<example>>☀☀☀

>>> import pandas as pd
>>> import numpy as np
>>> df = pd.DataFrame(np.arange(12).reshape(3, 4))
>>> df
   0  1   2   3
0  0  1   2   3
1  4  5   6   7
2  8  9  10  11
>>> df.columns = ['I', 'II', 'III', 'IV']
>>> df
   I  II  III  IV
0  0   1    2   3
1  4   5    6   7
2  8   9   10  11
>>> df.to_pickle(r"D:/aa/df")
 
>>> dff = pd.read_pickle(r"D:/aa/df")
>>> dff
   I  II  III  IV
0  0   1    2   3
1  4   5    6   7
2  8   9   10  11

3. sklearn library operation

After testing, the saved file size numpy.save()is consistent with the method.

☀☀☀<<example>>☀☀☀

Note: The new version of sklearn will report an error, just import joblib directly

Reference: (note the comment below) cannot import name joblib from sklearn.externals

from sklearn.externals import joblib
# 保存x
joblib.dump(x, 'x.pkl') 
# 加载x
x = joblib.load('x.pkl')

The new implementation method

import joblib
# 保存x
joblib.dump(x, 'x.pkl') 
# 加载x
x = joblib.load('x.pkl')

4. Pickle library operation

☀☀☀<<example>>☀☀☀

import pickle
  
# 存储变量的文件的名字
filename = 'shoplist.data'
# 初始化变量
shoplist = ['apple', 'mango', 'carrot']
# 以二进制写模式打开目标文件
f = open(filename, 'wb')
# 将变量存储到目标文件中区
pickle.dump(shoplist, f)
# 关闭文件
f.close()
# 删除变量
del shoplist
# 以二进制读模式打开目标文件
f = open(filename, 'rb')
# 将文件中的变量加载到当前工作区
storedlist = pickle.load(f)
print(storedlist)

5. Pytorch library operation

Model saving and loading method 1:

keep:

torch.save(model.state_dict(), mymodel.pth)    # 只保存模型权重参数，不保存模型结构

transfer:

model = My_model(*args, **kwargs)  #这里需要重构模型结构，My_model
model.load_state_dict(torch.load(mymodel.pth))  #这里根据模型结构，调用存储的模型参数
model.eval()

Model saving and loading method two:

keep:

torch.save(model, mymodel.pth)  # 保存整个 model 的状态

transfer:

model=torch.load(mymodel.pth)  # 这里已经不需要重构模型结构了，直接 load 就可以  
model.eval()