%autosave 30

Autosaving every 30 seconds

import numpy as np

a = np.arange(16).reshape(4, 4)

a

array([[ 0,  1,  2,  3],
       [ 4,  5,  6,  7],
       [ 8,  9, 10, 11],
       [12, 13, 14, 15]])

np.save('a.npy', a)

!cat a.npy

�NUMPY
v{'descr': '<i8', 'fortran_order': False, 'shape': (4, 4), }                                                          

	


np.savez('a.npz', a, a * 2)

!cat a.npz

PK!����

	arr_0.npy


�NUMPY
v{'descr': '<i8', 'fortran_order': False, 'shape': (4, 4), }                                                          

	
PK!o�d

	arr_1.npy


�NUMPY
v{'descr': '<i8', 'fortran_order': False, 'shape': (4, 4), }                                                          

PK
!����

	�
arr_0.npyPK
!o�d

	�
;
arr_1.npyPKnv

x = np.memmap('a.dat', mode='write', shape=(4,4), dtype=np.float64)
x[:] = 1
del x

with open('a.dat', 'rb') as fh:
    number = fh.read(8)
np.frombuffer(number)

array([1.])

import pickle

data = [1, 2.0, (3.0, 'Hello')]

with open('data.dat', 'wb') as fh:
    pickle.dump(data, fh)

!head data.dat

�]q(K
G@G@XHelloq
�qe.

with open('data.dat', 'rb') as fh:
    loaded = pickle.load(fh)
    print(loaded)
    assert loaded == data

[1, 2.0, (3.0, 'Hello')]

s = pickle.dumps(data)

loaded = pickle.loads(s)
print(loaded)
assert loaded == data

[1, 2.0, (3.0, 'Hello')]

from joblib import Memory

memory = Memory(location='cache')

@memory.cache
def double(x):
    return x * 2


print(double(1))
print(double(2))

________________________________________________________________________________
[Memory] Calling __main__--home-jovyan-work-__ipython-input__.double...
double(1)
___________________________________________________________double - 0.0s, 0.0min
2
________________________________________________________________________________
[Memory] Calling __main__--home-jovyan-work-__ipython-input__.double...
double(2)
___________________________________________________________double - 0.0s, 0.0min
4

import os

os.listdir('cache/joblib/__main__--home-jovyan-work-__ipython-input__/double/d3ffa92536e9b2aeb96c6d0e11ccd857')

['output.pkl', 'metadata.json']